54east / Insights / AI transformation vs digital transformation

AI transformation vs digital transformation · Insight

AI transformation is not digital transformation, sped up

Demand is real, and faster than the last decade. Value is not following at the same speed. The market is selling seats, slides, and software.

54east · Abu Dhabi · Updated 2026-09-17 · Open access

The answer: AI transformation is in demand because the tools are real. Adoption scaled in about three years. Attributed earnings did not. Hire a seat, a catalogue, a product, a shop, or a build — not one as if it were another.

---

The last decade put work on a screen. This one puts judgment on a screen.

Digital transformation, as it was actually practised from about 2012 to 2022, had a familiar shape. Move the process onto software. Stand up a cloud. Buy an ERP, a CRM, an analytics stack. Hire a systems integrator. Run a change programme so people would log into the new system instead of the spreadsheet. It took years. It cost what a capital programme costs. When it worked, the organisation could see the same work the same way, from more than one office.

AI transformation is being sold as that motion, compressed. It is not.

The last decade digitised records and routes. This one is trying to digitise judgment — the part of the work that used to live in a person: which inquiry is serious, which mix is legal, which correspondence can leave the building, which exception is real. A chatbot that drafts is not an ERP that posts. One produces a suggestion. The other produces a fact in the ledger. Treating them as the same “transformation” is how boards approve spend they cannot later find on the P&L.

The demand is not invented. The speed is not invented. The confusion between the two eras is.

Demand, as the numbers actually stand

Consumer adoption ran ahead of every prior consumer-internet curve. UBS, using Similarweb, estimated ChatGPT at 100 million monthly active users in January 2023 — two months after launch. TikTok took about nine months to the same mark; Instagram about two and a half years.^ubs-chatgpt

Stanford HAI’s 2026 AI Index puts generative-AI population adoption at 53 percent within three years, faster than the personal computer or the internet. The UAE is one of the countries running ahead of what income would predict, at 54 percent. The United States, which still leads in private AI capital, ranks 24th on that consumer measure, at 28.3 percent.^{stanford-2026 stanford-news}

Capital followed. Global corporate AI investment reached $581.7 billion in 2025, up 130 percent on the year. Private investment was $344.7 billion. Generative-AI companies took $170.9 billion of that private total — nearly half, and more than 200 percent up on 2024.^{stanford-2026}

Enterprise spend is on a similar slope. IDC put worldwide AI spending at $235 billion in 2024 and forecast $632 billion in 2028, a 29 percent compound annual growth rate. A later IDC forecast, driven by agentic systems, takes AI IT spending to $1.3 trillion in 2029.^{idc-632 idc-13}

That is still a slice of the digital-transformation pile, not a replacement for it. IDC has digital-transformation investment approaching $4 trillion by 2027–2028, about 70 percent of ICT spend, with AI recently around 17 percent of that DX total.^idc-dx AI is eating a growing share of a still-gigantic digital budget. It has not made the previous decade’s systems irrelevant. It is landing on top of them.

Inside companies, “we use AI” is now the default answer. McKinsey’s Global Survey tracked AI use in at least one business function from 55 percent, to 72 percent, to 78 percent, to 88 percent in the 2025 survey — and 89 percent in the 2026 survey.^{mck-rewire mck-2025 mck-2026} Generative AI specifically went from roughly a third of organisations in 2023 to 65 percent in early 2024, 71 percent later in 2024, and 88 percent reporting regular organisational AI use a year after that.^{mck-2024 mck-rewire}

The Gulf is not a lagging market on that chart. The UAE’s National Strategy for Artificial Intelligence 2031 set a public ambition — world leadership in AI, aligned to Centennial 2071, with up to AED 335 billion in extra growth as the strategy’s own figure.^uae-2031 Oxford Insights’ Government AI Readiness Index 2025 ranks the UAE 19th of 195, with a public-sector adoption score among the highest in the table. In November 2025 the Minister of State for Artificial Intelligence put utilisation of AI tools across government entities at 97 percent.^{oxford wam-97} Those figures, and the execution gap underneath them, are set out in the companion paper AI readiness before you build. They are not repeated here as a second maturity-model poster.

Demand, then, is not the question. Almost everyone has started. Almost no one has finished.

Scale without value is the actual story

The 2010s taught a generation of executives that “transformation” could be declared before it had paid. McKinsey’s 2018 Global Survey on digital transformations found that only 16 percent of respondents said their digital transformations had both improved performance and equipped the organisation to sustain those gains. Another 7 percent improved performance and then lost it. In oil and gas, automotive, infrastructure and pharmaceuticals the success band was 4 to 11 percent.^mck-dx-2018

That is the number to carry, not the orphan “70 percent of digital transformations fail,” which is a change-management round number with a weak empirical trail. Quote McKinsey’s digital survey as it stands.^mck-dx-2018

The AI era is repeating the shape, faster.

McKinsey’s 2025 survey: 88 percent using AI in at least one function; the majority still experimenting or piloting; about one-third beginning to scale. Twenty-three percent scaling an agent somewhere; in any given function, no more than 10 percent.^mck-2025

McKinsey’s 2026 survey (1,719 respondents, 4 May–8 June 2026): 89 percent using AI; 44 percent saying it is scaling across the enterprise, up from 38 percent; 56 percent using it in three or more functions. The share attributing a positive EBIT contribution to AI: 37 percent — essentially unchanged from 2025. The share McKinsey calls high performers, attributing at least 5 percent of EBIT to AI and describing the value as significant: 6 percent, also unchanged. Individual productivity is widely reported. The enterprise number is not moving with adoption.^mck-2026

Gartner, reviewing implementations through the end of 2025, reported that at least half of generative-AI projects were abandoned after proof of concept — data quality, risk controls, cost, or unclear value. An earlier Gartner forecast had been 30 percent. The later reading was worse.^gartner-half

MIT NANDA’s 2025 GenAI Divide found 95 percent of organisations getting zero return as the authors defined it: no marked, sustained productivity or P&L effect. That is a P&L test, not a demo test. It is already on the record in AI readiness before you build. Do not flatten it into “AI doesn’t work.”^nanda

The honest sentence is narrower: adoption scaled in about three years. Attributed earnings did not. Digital transformation took a decade to teach that lesson. AI is teaching it before most programmes have named an owner.

What is actually different from the digital decade

Two differences do the damage if they are ignored.

Probabilistic output in a deterministic institution. An ERP posts or it does not. A model is useful and occasionally wrong in the same afternoon. Ministries, banks, hospitals and plants are built for the first kind of error, not the second. Governance is not a slide at the end of the digital programme. It is the operating model. That is the argument of The AI governance gap.

Speed of the demo, slowness of the workflow. Digital programmes failed slowly enough that a steering committee could notice. Generative tools produce a convincing artefact on day one. The organisation then spends a year discovering that the artefact is not connected to SIMA, or SAP, or the correspondence archive, or the person who is allowed to sign. The demo hid the integration. Digital transformation at least made the integration the project.

A third difference is political, and specific to this region. Digital transformation in the Gulf often meant buying a global platform and running it in a regional data centre. AI transformation, if it touches citizen files, well data, or official correspondence, collides with jurisdiction — keys, compulsion, support access, exit — not just a region code. That collision is Why a cloud region isn’t a jurisdiction. It did not exist in the same form when the question was “do we have Salesforce.”

The options on the market, without the brochure

Four kinds of firm will take the meeting. They end in four different places. 54east’s public comparison table names them as strategy consultancy, software product, development firm, and a builder that keeps strategy and handover in the same team.^compare The market in 2026 is a little wider than that table, because a fifth option — the seat — now arrives before any of them are hired.

1. The seat. Microsoft 365 Copilot, Google Gemini for Workspace, in-app assistants inside the CRM and the design suite. Commodity capability, bought as a licence. Correct for summarising, drafting, and meeting notes that do not leave a regulated boundary. Wrong as a transformation programme. A seat does not know the institution’s approval chain. It does not write into the system of record unless someone builds that, which is no longer a seat. Buy it. Do not pretend it is the strategy.

2. The catalogue. Global strategy houses and the AI practices attached to them. The deliverable is a ranked list of use cases, a target operating model, sometimes a “lighthouse.” This is the digital-decade motion with new nouns. It is useful for a board that has not yet decided whether. It is not a system. The failure mode is documented in Build, buy, or wait: a catalogue is a diagnostic, not a decision.

3. The product. Vertical software with an “AI” module: dispatch with a forecasting add-on, a correspondence suite with a drafting pane, an agent platform that promises any workflow. Buy when the workflow is generic and the hosting bar can be met. Do not buy a template and call the gap “configuration” when the gap is the institution’s actual process. Test two still applies: if classification, IA, or sector rules close the buy, the product is not available, however mature the demo.

4. The development shop. Capable engineers, usually offshore or multi-shore, who will build what is specified. Core strength is code. The gap is the specification. Digital-decade SI programmes failed here too: the integrator delivered the ticket, the business never owned the process. In AI the gap is sharper, because the specification has to include when the model is allowed to act, who reviews it, and what happens when it is wrong.

5. The in-country build. A team that will sit on the workflow, write the decision (build, buy, wait), put a system next to the tools already there, keep the original system of record untouched until it has earned a write, and leave an owner. This is slower to start than a seat and less spectacular than a catalogue. It is the only option that matches what the numbers above are actually measuring: not “is AI present,” but “did a named loop move.”

None of these five is illegitimate. The error is hiring one as if it were another. Seats are not operating models. Catalogues are not systems. Products are not your approval chain. Code shops are not governance. And an in-country build that starts without a scored readiness check is just a more expensive abandoned proof of concept.

A sixth option exists and is usually unspoken: wait, on purpose. Name the data gap, the owner gap, or the hosting bar. Put a date on it. That is a decision. It is under-used because it does not photograph well in a ministry annual report. It is how the 16 percent who sustained digital value, and the 6 percent McKinsey now calls AI high performers, actually start — by refusing to scale a demo.

Why this decade will feel faster than it is

Three mechanics make AI look like it is transforming the company before it has.

Employees already use the tools. Stanford: four in five US students use generative AI for schoolwork; professionals are picking up prompting without a programme. The UAE is among the countries where AI-engineering skills are being learned fastest.^{stanford-news} Shadow IT in the digital decade was a rogue Salesforce instance. Shadow IT now is a personal ChatGPT session over a customer file. Adoption happened. Control did not.

Vendors can ship a convincing surface in a sprint. The digital decade required a data model before a screen. Generative tools invert that. The screen comes first. The data model, the audit trail, and the stop switch are “phase two,” which is where Gartner’s abandoned proofs of concept live.

Boards have a political deadline. UAE entities are scored on utilisation. Listed companies are asked on earnings calls. None of that is a reason to skip the workflow. It is a reason the catalogue sells.

The organisations that will keep the value are doing something more boring, and more like the digital programmes that actually stuck: they pick one loop, they connect it to a system of record they already trust, they put a human on the output that can hurt them, and they measure a number a superintendent or a director can read on a Friday.

That is not a 54east statistic. It is the shape that survives contact with McKinsey’s EBIT split, Gartner’s PoC graveyard, and every plant that already has an ERP and a dispatch system that do not talk.

What to do with the comparison

If you ran a digital programme last decade, three habits transfer. Three do not.

Transfer. Executive sponsorship with a named owner, not a committee. Integration to the system of record as the project, not a sequel. Training as handover, not a webinar after go-live.

Do not transfer. The five-year roadmap with forty initiatives. The assumption that a configured product is the process. The belief that hosting in-region closed the sovereignty question.

Add, because this is not digital. A written rule for when a model may only recommend. A reconstructable trail for anything that touches a regulated or official file. A kill metric: if the weekly number does not exist by a named date, the programme does not open the next wave.

Build, buy, or wait remains the decision. The seat is usually a buy. The institutional workflow is usually a build, or a wait until the data and the owner exist. The catalogue is not on the list.

Own the loop, then own the system

AI transformation is in demand because the tools are real and the competitive clock is real. It is being adopted at a speed digital transformation never managed. It is producing attributed earnings for a small minority, and abandoned proofs of concept for a large one. That is not cynicism. It is the survey record, three years in.

The last decade’s lesson still holds, and it is cheaper to remember than to relearn: transformation is not the presence of a platform. It is a changed number on a named workflow, with an owner who is still there after the vendor leaves.

54east builds that loop with the team that has to live in it. The engagement ends in a written decision and, where the decision is build, a system the institution owns. If the current programme is a seat count, a catalogue, or a proof of concept with no Friday number, talk to our team. Built here. Owned by you.

---

Notes

1. Reuters, “ChatGPT sets record for fastest-growing user base - analyst note,” 1 February 2023, reporting a UBS note that used Similarweb data. Estimated 100 million monthly active users in January 2023, two months after the 30 November 2022 launch. TikTok ~9 months to 100 million; Instagram ~2.5 years (Sensor Tower, as cited). https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/
2. Stanford Institute for Human-Centered AI, Artificial Intelligence Index Report 2026, Economy chapter. Global corporate AI investment $581.7 billion in 2025 (+129.9%); private investment $344.7 billion (+127.5%); generative-AI private funding $170.9 billion. Organisational AI adoption 88% in 2025 (McKinsey series, as compiled). https://hai.stanford.edu/ai-index/2026-ai-index-report; chapter PDF: https://hai.stanford.edu/assets/files/aiindexreport2026chapter4economy.pdf
3. Shana Lynch, “Inside the AI Index: 12 Takeaways from the 2026 Report,” Stanford HAI. Generative AI 53% population adoption in three years; Singapore 61%, UAE 54%, United States 28.3% (24th). US consumer value of genAI tools $172 billion annually by early 2026. https://hai.stanford.edu/news/inside-the-ai-index-12-takeaways-from-the-2026-report
4. IDC, via Business Wire, 19 August 2024: worldwide AI spending (AI-enabled applications, infrastructure, related IT and business services) $235 billion in 2024, forecast $632 billion in 2028, 29.0% CAGR 2024–2028. https://www.businesswire.com/news/home/20240819177906/en/Worldwide-Spending-on-Artificial-Intelligence-Forecast-to-Reach-%24632-Billion-in-2028-According-to-a-New-IDC-Spending-Guide
5. IDC, 26 August 2025: AI IT spending growing 31.9% year-over-year 2025–2029, reaching $1.3 trillion in 2029, with agentic AI as the stated driver. https://my.idc.com/getdoc.jsp?containerId=prUS53765225
6. IDC Worldwide Digital Transformation Spending Guide, May 2024 update: DX spending almost $4 trillion in 2027, 16.2% CAGR 2022–2027. Later IDC commentary: DX approaching $4 trillion by 2028, ~70% of ICT spend; AI around 17% of DX spend at the time of that note. https://www.hpcwire.com/bigdatawire/this-just-in/idc-worldwide-spending-on-digital-transformation-is-forecast-to-reach-almost-4t-by-2027/; https://www.idc.com/resource-center/blog/navigating-digital-transformation-amid-economic-uncertainty/
7. McKinsey, “The state of AI in early 2024.” 65% of respondents said their organisations regularly used generative AI in at least one function, nearly double the prior survey ten months earlier. Fielded 22 February–5 March 2024, n=1,363. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2024
8. McKinsey, “The State of AI: How organizations are rewiring to capture value” (2025 PDF of the mid-cycle survey). AI use in at least one function: 78%, up from 72% in early 2024 and 55% a year earlier. Generative AI: 71%, up from 65% in early 2024. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value
9. McKinsey, “The State of AI: Global Survey 2025.” 88% regular AI use in at least one function, up from 78% a year earlier; majority still experimenting or piloting; approximately one-third beginning to scale. 23% scaling an agentic system in at least one function; 39% experimenting with agents; in any given function, ≤10% scaling agents. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
10. McKinsey Global Survey on the State of AI 2026, as reported 25 August 2026. Fielded 4 May–8 June 2026, n=1,719, 97 countries (methodology as carried by Magica). 89% regular AI use; 44% enterprise-wide scaling (from 38%); 56% in three or more functions; 37% attributing positive EBIT impact (flat on 2025); 6% “high performers” (≥5% of EBIT, value described as significant), also flat. Primary press: The Register, 25 August 2026. Methodology detail: Magica. Prefer McKinsey’s own page when it is posted at the same URL as the 2025 survey.
11. McKinsey, “Unlocking success in digital transformations,” 2018 Global Survey. 16% said digital transformations improved performance and equipped the organisation to sustain the change; 7% improved but did not sustain. Fewer than 30% succeed is McKinsey’s broader transformation research, not this digital survey’s measured failure rate. Traditional industries 4–11% success. https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/unlocking-success-in-digital-transformations
12. Gartner, “Why Half of GenAI Projects Fail”: at least 50% of generative AI projects abandoned after proof of concept by end-2025. July 2024 forecast had been 30% by end-2025. https://www.gartner.com/en/articles/genai-project-failure; https://www.gartner.com/en/newsroom/press-releases/2024-07-29-gartner-predicts-30-percent-of-generative-ai-projects-will-be-abandoned-after-proof-of-concept-by-end-of-2025
13. Aditya Challapally, Chris Pease, Ramesh Raskar and Pradyumna Chari, The GenAI Divide: State of AI in Business 2025, MIT NANDA. 95% of organisations getting zero return as defined (no marked, sustained productivity or P&L impact). Coverage: The Register, 18 August 2025
14. UAE National Strategy for Artificial Intelligence 2031. Vision of world leadership in AI by 2031; up to AED 335 billion extra growth as stated in the strategy. https://ai.gov.ae/wp-content/uploads/2021/07/UAE-National-Strategy-for-Artificial-Intelligence-2031.pdf
15. Oxford Insights, Government AI Readiness Index 2025 (corrected January 2026 file; 195 countries; new methodology). UAE rank 19. PDF: https://oxfordinsights.com/wp-content/uploads/2026/01/Government-AI-Readiness-Report-2025-1.pdf
16. WAM, 5 November 2025. Minister Omar Sultan Al Olama: 97% utilisation of AI tools across government entities; federal AI Readiness Index launched. https://www.wam.ae/en/article/bmkc6nl-artificial-intelligence-readiness-index-for
17. 54east, How we compare — strategy consultancy, software product, development firm, and a single team from decision through handover. https://54east.ai/compare/

Related whitepapers