The agent in the back office

Executive summary

ERP is not dead. The commercial model around it is. No large enterprise has decommissioned a tier-one HRIS or finance system in favour of an AI-native alternative, the frontier AI labs run Workday themselves, and UK and EU regulation is hardening around exactly the architecture the incumbents provide. The real movement is commercial: consumption-based pricing is quietly repricing the back office.

Four decisions matter for 2026 and 2027.

Platform bets. Push back-office work onto the ERP vendor and its ecosystem partners by default. Reserve specialist tools for genuinely differentiated, usually customer-facing, capabilities.

Consumption. Flex credits, AELA and Joule reprice the back office as metered consumption that neither you nor the vendor can reliably forecast. Model the cost envelope before signature, not after.

Liability. Mobley v Workday shows AI vendors can be directly liable for discriminatory screening outcomes, and buyers are not insulated. Audit your screening vendors and your claim to meaningful human involvement now.

Deadlines. EU AI Act Article 10 obligations apply from 2 August 2026, the FRC and ICO guidance is already in force, and SAP ECC 6.0 mainstream maintenance ends in December 2027.

The work of 2026 and 2027 is to govern, instrument and augment the system you already have. The cheapest first step is an integration optimisation review, done before any consumption contract is signed.

ERP is not dead. But the commercial model around it is.

In early 2026, Workday, SAP, Oracle, ServiceNow and Salesforce lost more than a trillion dollars of market value in three months. The phrase doing the rounds is the ‘SaaSpocalypse’.^{1 2}

The argument is that agentic AI will make enterprise software obsolete, and that the CFO who renews a Workday contract in 2026 will look like the IT director who renewed a mainframe lease in 1995.³

It’s a good story. But it’s not what we are seeing in the boardroom.

As of July 2026, we can find no Fortune 500 or FTSE 100 company that has decommissioned a tier-one HRIS or finance system in favour of an AI-native alternative.

The frontier AI labs themselves are Workday customers.

The regulatory regime in the UK and EU is formalising around exactly the architecture the incumbents already provide.

None of that means the existing model continues unchanged.

What is changing, fast, is the commercial relationship with the vendor. Workday’s flex credits, Salesforce’s AELA, SAP’s Joule consumption and Oracle’s Agent Studio pricing are repricing back-office software in ways most CFOs have not modelled, and most procurement teams cannot meaningfully negotiate. The serious question for the next eighteen months is not whether to rip out your system of record. It is how to govern, instrument and augment it under tightening regulation, while not signing a consumption contract that funds your vendor’s AI roadmap with an open-ended cheque from your P&L.

This paper sets out what we are seeing in CFO and CHRO conversations across our client base, and what to do about it.

Rumours of ERP’s death are greatly exaggerated

The narrative that ERP is dead is loudest from people with something to sell.

The first group is incumbent software CEOs trying to consolidate. Bill McDermott of ServiceNow framed agentic AI at Knowledge 2025 as the moment “the industrial software complex of the 20th century is being integrated, and in some cases, drastically reduced in numbers of applications.” ⁴

During the February 2026 sell-off, he told the Wall Street Journal, “We don’t live in a SaaS neighbourhood.” Marc Benioff went further on Salesforce's Q4 FY26 earnings call, arguing that companies that fail to cannibalise themselves will be cannibalised by competitors.

Having watched Salesforce shares fall around 40 per cent from peak, Benioff stopped trying to dismiss the “SaaSpocalypse” and claimed it.⁵

The second group is venture capital funding the next wave. Sarah Tavel laid the intellectual groundwork in 2024 with “AI startups: sell work, not software”,⁶ arguing that LLMs allow pricing against labour cost rather than per-seat. Sam Lessin announced he would short the structured CRM companies. Julien Bek of Sequoia, who wrote the Series A cheque to Rillet, was more candid:

“The general ledger is the beating heart of the finance function, and so asking a company to remove it is a kind of open-heart surgery.”Julien Bek, Sequoia Capital⁷

The third is one analyst house. Phil Fersht and HFS Research coined “Services-as-Software” in 2024 and project a 1.5 trillion-dollar market by 2035. The forecast methodology is not public. It is HFS’s own number, not Gartner’s or IDC’s. Seth Ravin of Rimini Street is the bluntest of all, declaring that “ERP software is dead.” Rimini sells third-party support, so it has a direct commercial reason to shrink the perceived value of the core platforms.^{8 9}

Two things are worth noticing about that list.

First, the actual evidence each offers is thinner than the rhetoric. Klarna is the central proof point, and on inspection, it collapses (more on this below). Anthropic is the second, and it collapses harder. Beyond those two, the AI-native challengers most often cited as evidence (Rillet, Campfire, DualEntry, Sierra, Decagon) are doing something real but in a specific market segment. Their named customer lists are growth-stage technology companies under roughly 50 million dollars of revenue: Windsurf, PostHog, Coder, Replit, Trust & Will, Flex, Advisor360. That is a genuine mid-market displacement story against NetSuite and Sage Intacct. It is not an enterprise replacement story against Workday, SAP S/4HANA or Oracle Fusion.^{10 11 12 13 14}

Second, the most credible voice on the rip-and-replace side is undermining his own pitch. Rujul Zaparde, CEO of Zip, the procurement vendor at a 2.2-billion-dollar valuation with named customers including T-Mobile, Mars, OpenAI, Anthropic, LinkedIn and Block, describes his own product as follows:

“Those systems [SAP, Oracle] are great systems, but they’re systems of record. Zip is the orchestration layer that sits on top of those systems.”Rujul Zaparde, CEO, Zip¹⁵

If the most-cited next-generation procurement vendor describes itself as a layer that sits on top, the rip-and-replace story is not what the rip-and-replace vendors are actually selling.

What the noise gets wrong

Three claims are doing a lot of work in the market. Each one collapses on inspection.

Counter-voices have been consistent. Aaron Levie said:

“Generally, once you have a business process, you want to be able to define that in, effectively, business logic with deterministic systems, just because the risk of that changing any given day is very high.”Aaron Levie, CEO, Box, TechCrunch Disrupt in October 2025²⁵

Josh Bersin has been direct for two years:

“I don’t think there’s an existential threat of people throwing out their HCM, CRM or ERP for ChatGPT. The AI is not going to build the system for you.”Josh Bersin, The Josh Bersin Company²⁶

McKinsey’s own briefing concludes flatly that AI agents are unlikely to replace the ERP in the near or medium term because of system complexity. Even Gartner, which authored the headline 450-billion-dollar agentic AI projection, published in June 2025, stated that more than 40 per cent of agentic AI projects will be cancelled by the end of 2027 on cost, value and risk grounds. Gartner analyst Anushree Verma coined the useful term “agent washing” for rebranded chatbots and RPA.^{27 28}

What is really happening

The system of record becomes a control plane

The architectural direction of travel is visible across every major back-office vendor. Workday’s Agent System of Record (now powered by Sana, after Workday’s 1.1 billion dollar acquisition in September 2025), SAP’s Joule Studio with its 400-plus embedded AI use cases, Oracle’s AI Agent Studio with 600-plus embedded agents, ServiceNow’s AI Agent Fabric, Microsoft’s Copilot for Finance bundled into M365: none of these vendors is sleepwalking through the agentic transition.^{29 30 31 32 33}

The pattern is the same across all of them. Agents themselves need a system of record, to be onboarded, permissioned, audited and retired. The HRIS and ERP pattern already provides that for employees. It is being extended to agents. Far from being abstracted away, the core is becoming a more central control plane (see Figure 1). Gartner’s Alastair Woolcock put it crisply:

“Vendors that embed AI within this control plane will shape workflow execution. Vendors that treat AI as an enhancement layer risk being abstracted.”Alastair Woolcock, Gartner³⁴

The reason this is happening, rather than a clean replacement, is technical and legal. A deterministic system produces the same output from the same input every time. A three-way match passes or fails. Posted journals reconcile. Payroll calculations reproduce. A probabilistic system, however capable, can produce different outputs on identical inputs.

That is a feature when extracting unstructured data from a messy invoice. It is a liability when the output is a journal entry, a payroll calculation, a statutory report or a hiring decision. Three chained agentic steps at 90 per cent reliability each give roughly 73 per cent end-to-end accuracy. That compounding is what finance and HR audit frameworks are not built to absorb.

The regulatory reinforcement is direct. The FRC’s Generative and Agentic AI guidance (March 2026) is the first from any audit regulator globally and makes clear that firms and Responsible Individuals remain accountable regardless of AI use. UK GDPR Article 22A, as amended by the Data (Use and Access) Act 2025, restricts solely automated decisions with legal or similarly significant effects, including pay, benefits, discipline, promotion, and termination. ³⁵

Article 10 of the EU AI Act categorises all HR recruitment, evaluation, promotion and termination AI as high-risk, with full obligations applying from 2 August 2026. Penalties reach 35 million euros or 7 per cent of global annual turnover. The ICO’s Recruitment Rewired guidance (March 2026) signals that “meaningful human involvement” claims will be tested closely. DORA, applied 17 January 2025, pulls critical third-party AI providers to EU financial entities into regulated scope.^{36 37 38 39 40}

Whatever else is true, the regulator will not accept a defence based on the assertion that the agent made the decision.

The system of record is where accountability lives.

What we are seeing in client conversations

The published case-study base for AI-on-ERP is thin and dominated by US firms and vendor-marketed material. Where we have direct visibility through our own client engagements, the picture is messier than any vendor deck suggests. We have anonymised the clients in the mini case studies below.

Client 1 is a Workday customer that we are helping roll the platform across the enterprise. The business positions itself to investors as AI-native and is pushing AI deep into operations, finance and HR. Six months ago, Client 1 committed to Workday Recruiting, with a price hold from Workday. They have since decided to jettison it in favour of Ashby, an AI-native Applicant Tracking System, on the basis that Ashby feels more current. The IT function disagrees, because consolidating onto fewer platforms was the original strategy. The fight is unresolved.

The point is not that Ashby is the wrong choice or that Workday Recruiting is the right one. The point is that the basic stakeholder alignment problem (IT versus the business, consolidation versus the latest thing, build versus buy) has not gone away because AI is now the variable. It has got worse. Client 1’s CIO and CHRO do not have a settled plan, because the market keeps moving under them. Workday changes its pricing model. Anthropic ships a new capability. Google announces something at Cloud Next. The plan from one Monday is obsolete by the next.

Client 2 went through a Sana Learn demonstration in which one attendee, the head of learning, audibly cried partway through. Her team of eight was responsible for learning content development, with translation outsourced to external agencies on six-week cycles. Sana Learn could plausibly halve the headcount and remove the translation supplier entirely. Her boss’s response was the right one: rather than take out the cost, redeploy the team to transform learning content for the wider organisation, and partner with operations to retrain call centre staff. The detail of how that gets funded is unresolved. But the strategic instinct, to use the productivity gain to do more rather than less, is the right one.

It is also the conversation most CFOs and CHROs are not yet having clearly with their boards, despite the data running the same way. Forrester reports that 55 per cent of employers regret AI-driven layoffs and 57 per cent of AI investment decision-makers expect headcount to increase, not decrease. BCG’s AI Radar 2025 found 68 per cent of companies expect to maintain workforce size. Mercer’s Global Talent Trends 2026, based on around 12,000 respondents, found 43 per cent expect no major change, 32 per cent foresee reduction and 13 per cent expect increase. The “50 per cent finance headcount cut” framing being pushed in some boardrooms is not in the data.^{42 43 44}

Client 3 illustrates the harder version of the same problem. The CHRO felt let down by Workday, which he had bought in 2017, expecting it to be the future, and was being approached by six or seven AI-native HR vendors offering to build solutions on top of Workday. Some were credible, some were not. The argument that ultimately landed was not that Workday’s AI was already best-in-class. It was that the smaller vendors were either going to be acquired by larger ones, lacked the capital to build out properly, or did not understand the HR function as deeply as Workday did. Trust the vendor that knows your function, not the vendor with the prettiest demo. Client 3 subsequently implemented Paradox, an AI recruiting solution that has since been acquired by Workday, thereby vindicating the exact consolidation logic that drove the decision.

Behind these three stories sits a structural issue every CFO and CHRO should recognise. CIOs and CTOs cannot finalise an agentic AI architecture, because the architecture they would commit to in any given month is invalidated by the next month’s vendor announcements. Workday reworked its AI commercial model repeatedly between September 2025 and mid-2026: flex credits unveiled at Workday Rising, Sana repositioned as the AI brand in March 2026 and, according to partner briefings, complimentary credits bundled with signature of the new subscription agreement by May.^{45 46} Its line on what partners may build has shifted over the same period.⁴⁷ That is one vendor. Multiply by the number of platforms in any large enterprise stack and the planning problem becomes obvious.

This calibrates the scepticism with which any ‘rip-and-replace’ pitch deserves to be heard. The history of major ERP change is not flattering. Lidl wrote off 600 million euros on SAP eLWIS between 2011 and 2018. Birmingham City Council’s Oracle Cloud migration ran to between 80 and 100 million pounds against a 5.3-million-pound annual budget; a Grant Thornton audit in February 2025 found the council had operated “without an adequate financial management system and cash receipting system for over two years.” Maine’s Workday HR project failed at 35 million dollars. Bain’s 2025 benchmarking survey of 480 IT leaders found that more than 80 per cent of ERP transformations miss budget, timeline and value goals. The baseline failure rate of major ERP change is already high. Adding agentic AI to a bad change programme will not improve it. It will make it worse, more expensively, and faster.^{48 49 50 51}

Mobley v Workday: the case you can’t ignore

Of all the AI-adjacent risks, Mobley v Workday goes most directly to the question CHROs should be asking.

Who carries the liability when an AI screening tool produces discriminatory outcomes?

The short answer the court has given so far is potentially the vendor, not just the employer. That is novel.

Derek Mobley, a Black man over 40 with anxiety and depression, applied to more than 100 jobs at companies using Workday’s AI-driven applicant screening from 2017 onwards. He was rejected every time, often within minutes, sometimes in the middle of the night, without human review. ⁵²

Four additional plaintiffs have joined his case, all over 40. In July 2024, Judge Rita Lin in the Northern District of California denied Workday’s motion to dismiss, accepting that Workday could be directly liable under Title VII, the ADEA and the ADA as an “agent” of the employers using its tools. Her reasoning is the line CHROs need to read:^{52 53 54}

“Workday’s software is not simply implementing in a rote way the criteria that employers set forth but is instead participating in the decision-making process by recommending some candidates to move forward and rejecting others.”Judge Rita F. Lin, Northern District of California⁵²

A software vendor whose tool recommends or rejects candidates is, for civil rights purposes, an agent of the employer.

In May 2025, the court granted preliminary certification as a nationwide collective action under the ADEA. Workday’s own filings disclosed that 1.1 billion applications were rejected using its software tools during the relevant period. The collective could include hundreds of millions of members. ⁵³

In July 2025 the scope expanded to include Workday’s HiredScore AI features, with Workday ordered to provide the court with a list of customer companies that had enabled HiredScore by 20 August 2025.^{54 55 56}

As of early July 2026, the case is in discovery on the merits: Workday’s further motion to dismiss was denied on 6 March 2026, and the opt-in window for the ADEA collective closed the following day with around 14,000 claimants.⁵⁷ No liability ruling has been issued. But the procedural precedent (see Figure 2) is itself the story. AI vendors can face direct liability under civil rights law, not just their customers. SaaS contractual disclaimers of the form “we only provide the tool, the employer decides” are being tested, and so far, not holding. The disparate-impact claim, which does not require proving intent, only outcome, has been allowed to proceed.

The candid view from clients is closer to a shrug than a serious risk assessment. The position we hear most often, summarised generously, is that algorithmic bias is the vendor’s problem to solve, not the buyer’s.

There is a commercial logic to the position. Better that Workday, with deeper pockets and deeper expertise, deals with explainability than an individual CHRO does. But it is also a hostage to fortune. An audit finding, for example, that no candidates with Irish names have been progressed for interview, is not going to be satisfied by “Workday made me do it.”

Two practical points for any CHRO whose name is on the customer list Workday was ordered to disclose.

The theory of liability applies to any AI screening vendor, not only Workday. Eightfold, HireVue, Paradox, Pymetrics, and any Applicant Tracking System with AI scoring are exposed to the same logic.

A separate case, Kistler v Eightfold, was filed in California in January 2026, naming PayPal in the litigation filings.⁵⁸ Harper v Sirius XM, filed in Michigan in August 2025 over the iCIMS screening tool, is progressing on similar theories.⁵⁹ The question is not whether your vendor is being sued.

The question is whether your contractual structure protects you when a similar action is filed against the next vendor on the list.

“Meaningful human involvement” is not a phrase your vendor’s compliance team can settle on your behalf. The ICO has signalled it will test the claim. A human rubber-stamping an algorithmic decision is not meaningful human involvement, by either the ICO’s reading or Judge Lin’s reasoning. If your screening process today consists of an automated rejection, optionally reviewed by a human who never overrides the algorithm, the safe default is that you cannot honestly claim the involvement is meaningful.

A few other recent AI failures put the risk in context.

Air Canada was held legally liable for a chatbot that invented a bereavement-fare policy in 2024; damages were modest, but the precedent was sharp.⁶⁰ McDonald’s ended its IBM drive-thru AI pilot in July 2024 after two years and 100-plus locations.⁶¹ Amazon’s recruiting AI was scrapped in 2018 after it penalised the word “women”. ⁶² iTutorGroup settled an EEOC case in 2023 for 365,000 dollars after its software auto-rejected women over 55 and men over 60.⁶³ The Deloitte Australia incident, a 290,000-dollar government report partially refunded in October 2025 after AI-generated errors, is the early-warning shot for professional services firms using AI in client deliverables.⁶⁴

Flex credits: the consumption contract you cannot meaningfully model

The single most important commercial development for any Workday, SAP, Salesforce or Oracle customer in the next eighteen months is the shift from per-seat to consumption-based pricing for AI-driven functionality.

Workday’s flex credits are the leading edge of this. Salesforce has shipped three Agentforce pricing models in under eighteen months: 2 dollars per conversation at launch, then Flex Credits at 10 cents per action, then per-user AELA licences at 125 dollars or more per user per month.^{65 66} Of the roughly 5,000 Agentforce deals Salesforce had announced by February 2025, reports suggested only around 3,000 were paid.⁶⁷ SAP and Oracle are following the same direction.

“Customers still adopt a cautious approach to investing in GenAI because of the unpredictability of pricing models.”Jan Cook, Senior licensing analyst, Gartner⁶⁸

The mechanics of Workday’s model, as set out at a recent partner briefing, deserve closer attention than most CFOs are giving them (see Figure 3).

Workday is moving all 11,000 customers onto a new commercial architecture with three components. Sana is the AI brand that replaces Illuminate. UMSA, the Universal Main Subscription Agreement, replaces the old MSA and crucially flips innovation from opt-in to opt-out, with a console exposing granular data contributions per AI feature. Flex credits is the consumption-based model funding all of this.⁶⁹

Each customer's subscription includes a platform entitlement: a defined threshold across three transaction types: API requests, integration events and document storage. Exceeding the threshold on any one triggers a charge; underuse on one type cannot offset overuse on another. Those overage charges are settled in flex credits, a virtual currency allocated to customers as part of their subscription. When the allocation runs out, additional flex credits must be purchased. There is no hard spend cap. Workday will not stop service when customers blow through entitlements.

Pricing is essentially fixed. Customers can negotiate the volume of pre-purchased credits and earn modest discounts on bulk, but it seems they cannot meaningfully negotiate the rate card. Customers cannot move credits from one bucket to another, so under-consumption in one area cannot offset over-consumption in another.

Three things follow from this for CFOs.

First, the salesperson on the other side of the table cannot accurately predict how many credits you will consume, and neither can you. You can estimate your transactions in finance and HR. You can probably estimate the business process calls those transactions make. You cannot reliably predict how many credits an agent will consume in handling them, because that depends on agent behaviour, fallback paths, retries and orchestration patterns that are still being built. Workday’s own usage estimator is reportedly not used internally with confidence.

The pattern is already visible at scale outside the ERP market. Uber gave its engineers Claude Code and burned through its entire 2026 budget for AI coding tools within four months, prompting its chief operating officer to question publicly whether the spend was producing more useful features.⁷⁰ A developer tool is not an ERP consumption model, but the dynamic is identical: metered AI usage at scale is structurally difficult to forecast, and the finance team is usually the last to know.

Second, the new entitlement model meters things that used to be included. Workday’s own disclosures show transaction volume growing far faster than revenue, from more than 800 billion transactions a year in September 2024 to around 1.4 trillion by early 2026, against annual revenue growth of around 13 per cent.^{71 72} The commercial subtext, almost-but-not-quite said by Workday on partner calls, is that flex credits exist to close that gap. Customers are, in effect, being asked to fund Workday’s AI investment through consumption charges on infrastructure that was previously bundled. This is not a criticism. It is a description. CFOs should price the commercial model accordingly.

Third, when a Sana Enterprise agent acts on data that lives outside Workday (Outlook, SAP, ServiceNow, Google Workspace) the API call goes to the other vendor. So far, those vendors have not introduced API metering for agentic traffic. They will. The day Microsoft, SAP, Oracle or Google start charging for agent-driven API calls is the day every customer’s three-year cost envelope shifts again. This is the iceberg under the water. SaaS vendors carry direct cost when other vendors’ agents hit their data centres, and that bill is going to land somewhere. Aaron Levie’s “100x more agents than people” point is the CFO’s real concern: per-seat unit economics break long before the platform itself does.

There is a clear practical first move here, and it is unglamorous. Most enterprise Workday installations have integrations pinging the platform every five minutes for data that has not changed. Under flex credits and the new UMSA console, that wasted traffic is now metered and visible. A focused integration optimisation review (looking at endpoint frequency, payload size, polling versus event-driven patterns, and the IPs and APIs hitting Workday) can take a meaningful percentage off the consumption bill before the agentic AI conversation has even started.

This is exactly the kind of unsexy operational work that the implementation-led parts of the consultancy market are not focused on, and where Workday-specialist partners can demonstrably move the dial.

Four big decisions for 2026/27

The serious advisory conversation for the next eighteen months comes down to four decisions.

None of them is “should we replace our ERP?”.

What works alongside the core

The operational picture

Where AI is genuinely working in production, it sits in three places: as a conversational or drafting layer over the core, as an exception-handling and reconciliation layer around the core, or as an orchestration layer above the core calling ERP APIs. The ERP or HRIS remains the book of record. The pattern is consistent across the disclosed evidence base.

In finance close and reconciliation, BlackLine’s Verity suite drafts variance narratives, reconciliation matches and AR outreach; the human approves; the posting lands in SAP, Oracle or Workday. CEO Owen Ryan’s framing is the right one for CFOs:

“CFOs need to leverage AI but remain personally liable for financial accuracy, so a “black box” solution is not an option.”Owen Ryan, CEO, BlackLine⁷⁵

AppZen claims Toyota recovered 700,000 dollars in duplicate expenses and policy violations through AI-assisted review layered over Concur, Workday Expenses, Oracle and SAP, with a customer list including Amazon, Airbus and Novartis.⁷⁶

In procurement and AP, Coupa named operational customer results in its Q4 FY26 disclosures, including a 60 per cent reduction in RFP preparation and up to 15 per cent savings for Xylem across 200 million dollars of sourcing. Zip processes 355 billion dollars of spend a year and lists T-Mobile, Mars, OpenAI, Anthropic, LinkedIn and Block among customers, all sitting on top of existing SAP Ariba, Oracle and Coupa estates.⁷⁷

In HR service desk and talent, Moveworks (now part of ServiceNow) has the most densely documented augmentation footprint, from Palo Alto Networks (351,000 hours saved) to Siemens (an AI assistant for 250,000 employees across 190 countries). New Relic handles PTO requests end-to-end in chat, with the transaction posting to Workday. The UI is abstracted; the system of record is not.⁷⁸

In financial services, HSBC with Google Cloud’s Dynamic Risk Assessment cut AML false positives by 60 per cent across around 900 million monthly transactions, adjacent to, not inside, the ERP.⁷⁹ Lloyds Banking Group has scaled Microsoft 365 Copilot to around 30,000 colleagues, reporting 93 per cent active use and around 46 minutes saved per person per day.⁸⁰ Barclays is rolling out Copilot to 100,000 colleagues, with HR, travel and policy queries handled by a ‘Colleague AI Agent’ while core HR data stays where it is.⁸¹ Unilever runs SAP as its system of record and layers AI around it, and by mid-2024 had put 150 projects through its formal AI assurance process, against an estate of more than 500 AI systems.⁸²

The pattern, repeatedly, is the same. The ERP and HRIS are the substrate. The AI is the layer.

HPE’s experience at Workday Rising 2024 makes the underlying point most clearly. Workday Skills Cloud, deployed three years earlier, was the substrate that made HPE’s later generative-AI HR ambitions possible. The unglamorous data work came first. There is a generation of CHROs and CFOs who will spend 2026 and 2027 learning that the AI investment they want to make in 2027 depends on the data and process work they should have started in 2025.⁸³

Workday is not going away, but will be different

Where Workday is the platform, our view is that the fastest credible route to value is solutions developed by Workday itself or its ecosystem partners, layered on top of the core. Sana Enterprise is being positioned as the orchestration layer across third-party apps (Outlook, Google Workspace, SAP, ServiceNow) as well as the AI brand inside Workday.

Whether Sana wins that orchestration role against Microsoft Copilot, Google Gemini and ServiceNow’s AI Agent Fabric is genuinely unresolved. Predicting the winner is not what Workday customers should be spending their time on. Preparing the core for whichever orchestrator wins is.

That preparation has three concrete components.

1. The first is integration optimisation, addressed above. Reduce wasted API traffic before the consumption clock starts.
2. The second is data and process readiness for agents: the skills cloud, the master data, the process design that determines whether an agent has clean inputs to work with or has to fight noise. The HPE story above makes the point: the unglamorous data work came first.
3. The third is governance, both for the data the UMSA console now exposes per AI feature, and for the agents themselves as they enter the Agent System of Record. This is where the Workday environment most clearly becomes a control plane: agents are onboarded, permissioned, audited and retired alongside employees, against the same data and the same access controls. Done properly, that gives a CFO and a CHRO a defensible position with the regulator. Done improperly, it gives them an exposure they did not know they had.

This is the work our Workday-specialist partner Preos is engaged in across the customer base. It is unglamorous, operationally specific, and it is where the difference between a 30 per cent flex-credits overage and a clean run is actually made.

The 70 per cent that consultants are ignoring

Three frameworks come up repeatedly in serious advisory conversations and are worth naming briefly.

MIT CISR’s Enterprise AI Maturity Model identifies four capabilities: Strategy, Systems, Synchronisation, Stewardship. It places financially-outperforming firms in stages 3 and 4. DBS Bank is their reference case: 350-plus AI use cases, AI economic impact from 150 million Singapore dollars in 2022 to 370 million in 2024, projected over a billion in 2025.⁸⁴

McKinsey’s “Rewired” six-capability model covers strategy, talent, operating model, technology, data, and adoption and scaling.⁸⁵

However, the most useful framing for the conversation a CFO and CHRO need to have is BCG’s 10-20-70 rule: 10 per cent algorithms, 20 per cent data and technology, 70 per cent people and processes (see Figure 4). BCG’s Build for the Future 2025 study of 1,250 enterprises found only 5 per cent are “future-built” and achieve transformative AI value at scale. The failure mode is well-documented: organisations over-invest in the 10 per cent (the model, the algorithm, the agent) and under-invest in the 70 per cent that determines whether the algorithm changes anything operationally.⁴¹

None of the three frameworks contemplates replacing the system of record. All frame AI as a re-layering of the operating model on top of it. The 70 per cent is where the embarrassments and the wins of 2026 and 2027 will both come from. It is also the part the implementation-focused parts of the consultancy market are systematically not focused on. That 70 per cent is where Change Associates works. Our business is the process, people and change work that the 10-20-70 research says determines whether the investment pays back: stakeholder alignment, operating model design, adoption, and the governance that makes a deployment defensible. The algorithms will keep improving without our help; the organisation will not.

Source Global Research’s Q2 2025 survey found 81 per cent of clients had paid for AI-related consulting support in the previous twelve months. The global technology consulting market is forecast to pass 400 billion dollars in 2026.

“Investing to upgrade their legacy infrastructure is at the top of [clients’] to-do lists. It is not something they are willing to hold off on until the market becomes clearer.”Nick Jotischky, Source Global Research⁸⁶

The market is moving. The question is whether the buyer ends up on the right side of the 10-20-70 split.

Conclusion: the conversation the boardroom should be having

The temptation to predict which agentic platform wins is strong, and worth resisting. The market will sort some of that out. Acquisitions will sort the rest. What will not sort itself out is whether your organisation has the operational discipline to make a small number of platform bets, model the consumption envelope honestly, govern the vendor liability that comes with AI in HR, and hit the regulatory deadlines.

The framing to close on is straightforward. What’s the task? What’s the problem? What’s the fit-for-purpose solution? Push as much as you can onto the ERP vendor that already knows your function. Don’t get blinded by the AI light. Don’t build it yourself.

The ERP is not dead. The commercial model around it is, and the work for CFOs, CHROs and CIOs in 2026 and 2027 is to govern, instrument and augment what they already have under tighter regulation, with better consumption modelling, and with a clearer view than they currently have of where their vendor liability sits.

The CFOs and CHROs who will be embarrassed in two years are not the ones still running Workday, SAP or Oracle. They are the ones who bought the agent without redesigning the process, signed the consumption contract without modelling the cost envelope, claimed meaningful human involvement in an automated decision the regulator will test, or find their company’s name on a notification list to several hundred thousand applicants who believe they were discriminated against.

That is the conversation the boardroom should be having now.

A practical place to start is the integration optimisation review described above: a bounded first engagement that trims the consumption bill before any agentic contract is signed and gives the board an honest view of where the estate stands. If that would be useful, we would be glad to talk.

References

• 1. Fortune, “AI agents from Anthropic and OpenAI aren’t killing SaaS” (10 February 2026). https://fortune.com/2026/02/10/ai-agents-anthropic-openai-arent-killing-saas-salesforce-servicenow-microsoft-workday-cant-sleep-easy/ (Accessed: 28/04/26)
• 2. TechCrunch, “SaaS in, SaaS out: here’s what’s driving the SaaSpocalypse” (1 March 2026), citing Aaron Holiday of 645 Ventures. https://techcrunch.com/2026/03/01/saas-in-saas-out-heres-whats-driving-the-saaspocalypse/ (Accessed: 28/04/26)
• 3. Gerstner, B. and Gurley, B., BG2 Podcast featuring Satya Nadella (19 December 2024). https://www.bg2pod.com/ (Accessed: 28/04/26)
• 4. Runtime, “ServiceNow CEO Bill McDermott: agentic AI puts IT back in control” (Knowledge 2025 coverage). https://www.runtime.news/servicenow-ceo-bill-mcdermott-agentic-ai-puts-it-back-in-control/ (Accessed: 28/04/26)
• 5. WebProNews, “Marc Benioff Declares the End of SaaS as We Know It” (26 February 2026). https://www.webpronews.com/marc-benioff-declares-the-end-of-saas-as-we-know-it-and-bets-salesforces-future-on-autonomous-ai-agents/ (Accessed: 28/04/26)
• 6. Tavel, S., “AI startups: sell work, not software” (March 2024). https://www.sarahtavel.com/p/a-few-sell-work-not-software-updated (Accessed: 28/04/26)
• 7. Crunchbase News, “Rillet Series B” (citing Julien Bek, Sequoia). https://news.crunchbase.com/fintech/startup-rillet-ai-seriesb-a16z-iconiq/ (Accessed: 28/04/26)
• 8. Fersht, P., HFS Research, “Services-as-Software: a 1.5 trillion dollar opportunity” (September 2024). https://www.horsesforsources.com/sas_1-5-trillion-dollar-opportunity_020725/ (Accessed: 28/04/26)
• 9. Diginomica, “ERP as we know it is dead: what’s next?” (Rimini Street analyst day coverage, December 2025). https://diginomica.com/erp-we-know-it-dead-whats-next-get-ready-agentic-ai-erp-says-rimini-street-cio-vijay-kumar (Accessed: 28/04/26)
• 10. Crunchbase News, Rillet Series B funding announcement. https://news.crunchbase.com/fintech/startup-rillet-ai-seriesb-a16z-iconiq/ (Accessed: 28/04/26)
• 11. Numeric, “Rillet vs Campfire: AI-native ERPs for high-growth companies” (buyer’s comparison). https://www.numeric.io/blog/rillet-vs-campfire (Accessed: 28/04/26)
• 12. PR Newswire, Campfire Series A announcement. https://www.prnewswire.com/news-releases/campfire-raises-35-million-series-a-led-by-accel-to-build-the-next-generation-ai-driven-erp-302494153.html (Accessed: 28/04/26)
• 13. CNBC, “Bret Taylor’s Sierra AI startup”. https://www.cnbc.com/2025/09/04/bret-taylor-sierra-ai-startup-salesforce-openai.html (Accessed: 28/04/26)
• 14. Business Wire, Decagon Series D announcement (January 2026). https://www.businesswire.com/news/home/20260128580542/en/ (Accessed: 28/04/26)
• 15. PYMNTS, “Zip says agentic procurement orchestration platform processed 355 billion dollars in spend” (citing Rujul Zaparde). https://www.pymnts.com/artificial-intelligence-2/2025/zip-says-agentic-procurement-orchestration-platform-processed-355-billion-in-spend/ (Accessed: 28/04/26)
• 16. CX Today, “Klarna didn’t replace Salesforce: it replaced them with alternative SaaS apps”. https://www.cxtoday.com/crm/klarna-didnt-replace-salesforce-it-replaced-them-with-alternative-saas-apps/ (Accessed: 28/04/26)
• 17. Diginomica, “Those shutting down Salesforce and Workday rumours from Klarna: no, we didn’t replace SaaS with LLM, admits CEO”. https://diginomica.com/those-shutting-down-salesforce-and-workday-rumors-klarna-no-we-didnt-replace-saas-llm-admits-ceo (Accessed: 28/04/26)
• 18. LaSoft, “Klarna walks back AI overhaul: rehires staff after customer service backlash”. https://lasoft.org/blog/klarna-walks-back-ai-overhaul-rehires-staff-after-customer-service-backlash/ (Accessed: 28/04/26)
• 19. MLQ.ai, “Klarna CEO admits aggressive AI job cuts went too far”. https://mlq.ai/news/klarna-ceo-admits-aggressive-ai-job-cuts-went-too-far-starts-hiring-again-after-us-ipo/ (Accessed: 28/04/26)
• 20. Ford, B., “Workday CEO says Anthropic and OpenAI use his company’s software”, Bloomberg (24 February 2026). https://www.bloomberg.com/news/articles/2026-02-24/workday-ceo-says-anthropic-and-openai-use-his-company-s-software (Accessed: 28/04/26)
• 21. Rachitsky, L., “Anthropic’s 1B to 19B growth run” with Amol Avasare, Lenny’s Newsletter (5 April 2026). https://www.lennysnewsletter.com/p/anthropics-1b-to-19b-growth-run (Accessed: 28/04/26)
• 22. Reworked, “Learning the wrong lessons from Workday losing its CTO to Anthropic”. https://www.reworked.co/employee-experience/workday-losing-its-cto-to-anthropic/ (Accessed: 28/04/26)
• 23. ERP Today, “Is Workday CTO’s exit to Anthropic signalling escalating AI pressure on HCM platforms?”. https://erp.today/is-workday-ctos-exit-to-anthropic-signaling-escalating-ai-pressure-on-hcm-platforms/ (Accessed: 28/04/26)
• 24. Battery Ventures, “First coding, next finance? AI adoption comes to the CFO suite” (1 April 2026). https://www.battery.com/blog/first-coding-next-finance-ai-adoption-comes-to-the-cfo-suite/ (Accessed: 05/07/26)
• 25. Yahoo Finance, “Box CEO Aaron Levie on AI”. https://finance.yahoo.com/news/box-ceo-aaron-levie-ai-184519642.html (Accessed: 28/04/26)
• 26. Bersin, J., “Is the tech market going nuts?” (September 2025). https://joshbersin.com/2025/09/is-the-tech-market-going-nuts/ (Accessed: 28/04/26)
• 27. McKinsey, “Bridging the great AI agent and ERP divide to unlock value at scale”. https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/bridging-the-great-ai-agent-and-erp-divide-to-unlock-value-at-scale (Accessed: 28/04/26)
• 28. Gartner, “Gartner predicts over 40 per cent of agentic AI projects will be cancelled by end of 2027” (25 June 2025). https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027 (Accessed: 28/04/26)
• 29. Workday, Agent System of Record product page. https://www.workday.com/en-us/artificial-intelligence/agent-system-of-record.html (Accessed: 28/04/26)
• 30. Diginomica, “Workday Rising 2025: does the acquisition of Sana mean Workday agrees SaaS is dead?”. https://diginomica.com/workday-rising-2025-does-acquisition-sana-mean-workday-agrees-saas-dead (Accessed: 28/04/26)
• 31. SAP News, “SAP connect business AI: new Joule agents and embedded intelligence” (October 2025). https://news.sap.com/2025/10/sap-connect-business-ai-new-joule-agents-embedded-intelligence/ (Accessed: 28/04/26)
• 32. Oracle, “Oracle introduces AI Agent Studio” (20 March 2025). https://www.oracle.com/news/announcement/oracle-introduces-ai-agent-studio-2025-03-20/ (Accessed: 28/04/26)
• 33. Microsoft, “Moving sales, service and finance to the frontier with Microsoft 365 Copilot” (10 September 2025). https://www.microsoft.com/en-us/microsoft-365/blog/2025/09/10/moving-sales-service-and-finance-to-the-frontier-with-microsoft-365-copilot/ (Accessed: 28/04/26)
• 34. Gartner, “Gartner expects most enterprises to abandon assistive AI for outcome-focused workflow by 2028” (2 April 2026), quoting VP Analyst Alastair Woolcock. https://www.gartner.com/en/newsroom/press-releases/2026-04-02-gartner-expects-most-enterprises-to-abandon-assistive-ai-for-outcome-focused-workflow-by-2028 (Accessed: 25/06/26)
• 35. Financial Reporting Council, “Innovative new guidance supports audit firm adoption of emerging AI technologies” (March 2026). https://www.frc.org.uk/news-and-events/news/2026/03/innovative-new-guidance-supports-audit-firm-adoption-of-emerging-ai-technologies/ (Accessed: 28/04/26)
• 36. ICO, “Automated decisions can streamline the hiring process with the right safeguards in place: Recruitment Rewired guidance” (31 March 2026). https://ico.org.uk/about-the-ico/media-centre/news-and-blogs/2026/03/automated-decisions-can-streamline-the-hiring-process-with-the-right-safeguards-in-place/ (Accessed: 28/04/26)
• 37. DLA Piper, “UK ICO report on automated decision-making in recruitment” (April 2026). https://privacymatters.dlapiper.com/2026/04/uk-ico-report-on-automated-decision-making-in-recruitment/ (Accessed: 28/04/26)
• 38. EU AI Act, official text and implementation timetable. https://artificialintelligenceact.eu/ (Accessed: 28/04/26)
• 39. Crowell & Moring, “Artificial intelligence and human resources in the EU: a 2026 legal overview”. https://www.crowell.com/en/insights/client-alerts/artificial-intelligence-and-human-resources-in-the-eu-a-2026-legal-overview (Accessed: 28/04/26)
• 40. EY Switzerland, “DORA: a new era of digital operational resilience”. https://www.ey.com/en_ch/insights/cybersecurity/dora-a-new-era-of-digital-operational-resilience (Accessed: 28/04/26)
• 41. Boston Consulting Group, “The Widening AI Value Gap: Build for the Future 2025” (September 2025). https://www.bcg.com/publications/2025/are-you-generating-value-from-ai-the-widening-gap (Accessed: 05/07/26)
• 42. Boston Consulting Group, “From Potential to Profit: Closing the AI Impact Gap” (AI Radar 2025, January 2025). https://www.bcg.com/publications/2025/closing-the-ai-impact-gap (Accessed: 25/06/26)
• 43. Mercer, Global Talent Trends 2026 report. https://www.mercer.com/about/newsroom/mercer-s-global-talent-trends-2026-report/ (Accessed: 28/04/26)
• 44. The Register, “AI layoffs to backfire: half quietly rehired at lower pay” (29 October 2025), citing Forrester, Predictions 2026: The Future of Work. https://www.theregister.com/2025/10/29/forrester_ai_rehiring/ (Accessed: 05/07/26)
• 45. Workday Newsroom, “Workday Illuminate expands with new AI agents for HR, finance, and industry” (16 September 2025), announcing flex credits. https://newsroom.workday.com/2025-09-16-Workday-Illuminate-TM-Expands-with-New-AI-Agents-for-HR,-Finance,-and-Industry (Accessed: 05/07/26)
• 46. Workday Newsroom, “Introducing Sana from Workday” (17 March 2026). https://newsroom.workday.com/2026-03-17-Introducing-Sana-from-Workday-Superintelligence-for-Work-That-Finds-Answers,-Takes-Action,-and-Automates-Workflows (Accessed: 05/07/26)
• 47. ERP Today, “Gatekeeping enterprise AI? Workday redraws the build-vs-buy line for AI agents” (2 July 2026). https://erp.today/workday-ai-agents-build-vs-buy-governance (Accessed: 05/07/26)
• 48. The Chain, “Lessons from Lidl’s 600 million euro SAP failure”. https://www.thechain.media/p/lessons-from-lidls-600m-sap-failure (Accessed: 28/04/26)
• 49. Computer Weekly, “Birmingham City Council’s Oracle implementation explained: what went wrong”. https://www.computerweekly.com/news/366572935/Birmingham-City-Councils-Oracle-implementation-explained-What-went-wrong (Accessed: 28/04/26)
• 50. Governing, “Maine’s 35 million dollar human resources system failed: here’s why”. https://www.governing.com/now/maines-35-million-human-resources-system-failed-heres-why (Accessed: 28/04/26)
• 51. Bain & Company, “Is agentic AI the inflection point for scaling ERP transformations?” (2025). https://www.bain.com/insights/is-agentic-ai-the-inflection-point-for-scaling-ERP-transformations/ (Accessed: 28/04/26)
• 52. Seyfarth Shaw, “Mobley v Workday: court holds AI service providers could be directly liable for employment discrimination under agent theory” (July 2024). https://www.seyfarth.com/news-insights/mobley-v-workday-court-holds-ai-service-providers-could-be-directly-liable-for-employment-discrimination-under-agent-theory.html (Accessed: 28/04/26)
• 53. Proskauer Rose LLP, “AI bias lawsuit against Workday reaches next stage as court grants conditional certification of ADEA claim” (May 2025). https://www.lawandtheworkplace.com/2025/06/ai-bias-lawsuit-against-workday-reaches-next-stage-as-court-grants-conditional-certification-of-adea-claim/ (Accessed: 28/04/26)
• 54. Fisher Phillips, “Discrimination lawsuit over Workday’s AI hiring tools can proceed as class action: 6 things”. https://www.fisherphillips.com/en/insights/insights/discrimination-lawsuit-over-workdays-ai-hiring-tools-can-proceed-as-class-action-6-things (Accessed: 28/04/26)
• 55. SocialTalent, “The Workday lawsuit just changed everything: AI hiring audit”. https://www.socialtalent.com/blog/technology/workday-lawsuit-ai-hiring-audit (Accessed: 28/04/26)
• 56. FairNow, Workday lawsuit case tracker. https://fairnow.ai/workday-lawsuit-resume-screening/ (Accessed: 28/04/26)
• 57. Forbes, “A federal judge, a 1967 law and a billion rejected job applications” (29 May 2026). https://www.forbes.com/sites/sheilacallaham/2026/05/29/a-federal-judge-a-1967-law-and-a-billion-rejected-job-applications/ (Accessed: 05/07/26)
• 58. HR Brew, “AI-powered hiring platform Eightfold AI faces lawsuit over hiring data used to rate candidates” (29 January 2026). https://www.hr-brew.com/stories/2026/01/29/ai-powered-hiring-platform-eightfold-ai-faces-lawsuit-over-hiring-data-used-to-rate-candidates (Accessed: 28/04/26)
• 59. Fisher Phillips, “Another employer faces AI hiring bias lawsuit: 10 actions you can take to prevent AI litigation” (15 August 2025), on Harper v Sirius XM Radio, LLC (E.D. Michigan, filed 4 August 2025). https://www.fisherphillips.com/en/insights/insights/another-employer-faces-ai-hiring-bias-lawsuit (Accessed: 05/07/26)
• 60. McCarthy Tétrault, “Moffatt v Air Canada: misrepresentation by AI chatbot” (2024). https://www.mccarthy.ca/en/insights/blogs/techlex/moffatt-v-air-canada-misrepresentation-ai-chatbot (Accessed: 28/04/26)
• 61. Restaurant Business Online, “McDonald’s ending its drive-thru AI test” (July 2024). https://www.restaurantbusinessonline.com/technology/mcdonalds-ending-its-drive-thru-ai-test (Accessed: 28/04/26)
• 62. MIT Technology Review, “Amazon ditched AI recruitment software because it was biased against women” (10 October 2018). https://www.technologyreview.com/2018/10/10/139858/amazon-ditched-ai-recruitment-software-because-it-was-biased-against-women/ (Accessed: 28/04/26)
• 63. Berkshire Associates, “EEOC settles first AI bias lawsuit” (iTutorGroup, 2023). https://www.berkshireassociates.com/blog/eeoc-settles-first-ai-bias-lawsuit (Accessed: 28/04/26)
• 64. Australian Associated Press via The Washington Post, “Deloitte to partially refund Australian government for report with apparent AI-generated errors” (7 October 2025). https://www.washingtonpost.com/business/2025/10/07/australia-ai-errors-deloitte/01546f26-a34d-11f0-a79e-ccb5b1f59130_story.html (Accessed: 05/07/26)
• 65. The Register, “Salesforce pricing per AI conversation” (29 August 2024). https://www.theregister.com/2024/08/29/salesforce_pricing_per_ai_conversation/ (Accessed: 05/07/26)
• 66. SaaStr, “Salesforce now has 3+ pricing models for Agentforce” (2025). https://www.saastr.com/salesforce-now-has-3-pricing-models-for-agentforce-and-maybe-right-now-thats-the-way-to-do-it/ (Accessed: 05/07/26)
• 67. Salesforce Ben, “Why Agentforce adoption is slower than expected” (11 April 2025), citing The Information. https://www.salesforceben.com/why-agentforce-adoption-is-slower-than-expected-and-what-salesforce-needs-to-do/ (Accessed: 05/07/26)
• 68. The Register, “Salesforce opts for seat-based AI licensing as customers demand predictability” (12 December 2025), quoting Gartner senior licensing analyst Jan Cook. https://www.theregister.com/software/2025/12/12/salesforce-opts-for-seat-based-ai-licensing/ (Accessed: 05/07/26)
• 69. Workday, AI Flex Credits product page and Universal Main Subscription Agreement (UMSA) FAQ. https://www.workday.com/en-us/artificial-intelligence/ai-flex-credits.html and https://www.workday.com/en-us/legal/universal-contract-terms-and-conditions/umsa/faq.html (Accessed: 05/07/26)
• 70. Fortune, “Uber burned through its entire 2026 AI budget in four months” (26 May 2026). https://fortune.com/2026/05/26/uber-coo-ai-spending-tokens-claude-code/ (Accessed: 05/07/26)
• 71. PR Newswire, “Announcing Workday Illuminate: the next generation of Workday AI” (17 September 2024), citing more than 800 billion transactions a year. https://www.prnewswire.com/news-releases/announcing-workday-illuminate-the-next-generation-of-workday-ai-302249986.html (Accessed: 05/07/26)
• 72. Workday, “Workday announces fiscal 2026 fourth quarter and full year financial results” (February 2026). https://investor.workday.com/news-and-events/press-releases/news-details/2026/Workday-Announces-Fiscal-2026-Fourth-Quarter-and-Full-Year-Financial-Results/default.aspx (Accessed: 05/07/26)
• 73. Massachusetts Institute of Technology, NANDA initiative, “State of AI in Business 2025”. https://www.media.mit.edu/projects/nanda/ (Accessed: 28/04/26)
• 74. McKinsey, “The state of AI: how organisations are rewiring to capture value” (November 2025). https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai (Accessed: 28/04/26)
• 75. IT Brief, “BlackLine launches AI financial ops model for CFOs” (Verity suite). https://itbrief.asia/story/blackline-launches-ai-financial-ops-model-for-cfos (Accessed: 28/04/26)
• 76. AppZen / Financial Technology Report, “AppZen brings AI automation to expense and invoice management”. https://thefinancialtechnologyreport.com/appzen-bring-ai-automation-to-expense-and-invoice-management/ (Accessed: 28/04/26)
• 77. Shashi, “Coupa just raised the bar on AI procurement” (March 2026). https://www.shashi.co/2026/03/coupa-just-raised-bar-on-ai-procurement.html (Accessed: 28/04/26)
• 78. Moveworks, customer case studies. https://www.moveworks.com/us/en/customers (Accessed: 28/04/26)
• 79. SmartDev, “AI use cases in AML” (HSBC and Google Cloud Dynamic Risk Assessment). https://smartdev.com/ai-use-cases-in-aml/ (Accessed: 28/04/26)
• 80. Lloyds Banking Group, “Lloyds Banking Group scales adoption of Microsoft 365 Copilot” (21 October 2025). https://www.lloydsbankinggroup.com/media/press-releases/2025/lloyds-banking-group-2025/lloyds-scales-adoption-of-copilot.html (Accessed: 05/07/26)
• 81. Microsoft UK Stories, “Barclays to roll out Microsoft 365 Copilot to 100,000 colleagues” (9 June 2025). https://ukstories.microsoft.com/features/barclays-rolls-out-microsoft-365-copilot-to-100000-colleagues/ (Accessed: 05/07/26)
• 82. Unilever, “The EU AI Act has arrived: how Unilever is preparing” (30 July 2024). https://www.unilever.com/news/news-search/2024/the-eu-ai-act-has-arrived-how-unilever-is-preparing/ (Accessed: 05/07/26)
• 83. Diginomica, “Workday Rising 2024: customers reveal early lessons on gen AI in action”. https://diginomica.com/workday-rising-2024-customers-reveal-early-lessons-gen-ai-action (Accessed: 28/04/26)
• 84. MIT CISR, Weill, P., Woerner, S. and Sebastian, I., “Enterprise AI Maturity Model” (updated August 2025). https://cisr.mit.edu/ (Accessed: 28/04/26)
• 85. McKinsey & Company, “Rewired to outcompete” (the six capabilities of digital and AI transformation). https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/rewired-to-outcompete (Accessed: 25/06/26)
• 86. Consultancy.uk, “Global technology consulting market to surpass 400 billion dollars in 2026” (19 December 2025), citing Source Global Research; and Source Global Research, “Understanding the market for AI-related consulting services” (Q2 2025). https://www.consultancy.uk/news/42532/global-technology-consulting-market-to-surpass-400-billion-in-2026 and https://sourceglobalresearch.com/blog-post/understanding-the-market-for-ai-related-consulting-services (Accessed: 05/07/26)