Numbers and measurement

CEOs and AI in 2026: who gets the return, and what it means for healthcare leaders

Most companies use AI, and few see it in both revenue and cost. What the 2026 surveys say, what that small group does differently, and how a healthcare leader can apply it.

1 November 2026 · 12 min read

CEO surveys in 2026 show that most companies use AI and few see a return from it. In PwC's Global CEO Survey of January 2026, 12% of chief executives saw AI lift revenue and cut costs together. That small group tied AI to specific processes and measured numbers, such as reply time and cost per booking.

I am writing this for the director of a medical group who approved two or three AI subscriptions this year and was then asked by the board what they returned. Much larger companies face the same question, and the surveys give a clear picture of the answer.

What do the 2026 surveys say about AI return on investment?

The picture is consistent across studies: use is wide, and the financial effect is limited and concentrated in a small number of companies. These are the main figures, each with its source and date.

PwC Global CEO Survey, January 2026

PwC's 29th survey covered 4,454 chief executives in 95 countries and was released at Davos. On the previous twelve months:

  • 56% saw neither higher revenue nor lower costs from AI.
  • 30% saw higher revenue, and 26% saw lower costs.
  • 12% saw both.

The survey also recorded a difference: those who saw both were two to three times more likely to have embedded AI extensively in their products, in demand generation and in decision-making.

The Middle East edition of the same survey

In the regional results, covering more than 300 chief executives, 70% said they had a defined AI roadmap, against 51% globally. 39% in the Middle East and 43% in the Gulf said they used AI extensively in sales, marketing and customer service, against 22% globally.

The region is ahead on plans and on use in marketing and customer service. The question of return is the same.

NBER working paper, October 2026

Working paper 34836 from the US National Bureau of Economic Research, revised in October 2026, gathered answers from nearly 9,000 CFOs, CEOs and senior executives in seven countries, including the US, the UK, Germany and South Africa, in summer 2026. Its main findings:

  • 78% of firms use AI.
  • 80% of executives use it regularly, for 2.1 hours a week on average.
  • Executives estimate it raised labour productivity growth by about 0.2% a year over the last three years, and expect 0.6% a year over the next three.
  • 2% of firms account for more than half of all AI spending.
  • There was essentially no net effect on employment, alongside a lot of movement between roles.

This is an academic paper with a large sample, so we give it more weight than consultancy surveys.

The 2025 studies that came before

Three studies from 2025 say the same thing with different numbers:

  • McKinsey, November 2025: 88% of organisations use AI in at least one function, and 39% reported any impact on operating profit, mostly under 5%. About 6% were classed as high performers. The sample was 1,993 respondents in 105 countries.
  • Boston Consulting Group, September 2025: 5% of companies get value from AI at scale, and 60% reported minimal revenue and cost gains. The sample was 1,250 executives.
  • IBM, May 2025: 25% of AI initiatives delivered the expected return, and 16% were scaled across the company. The sample was 2,000 chief executives in 33 countries.

And in healthcare

The available healthcare figures are American, so read them as a sign of direction. In the American Medical Association survey published in March 2026, covering 1,692 physicians, more than 80% said they use AI in their work, up from 38% in 2023. Menlo Ventures found in October 2025 that 22% of US healthcare organisations had implemented AI tools built for the sector.

Why do so few companies see the return?

Our reading of the figures is that the gap comes from how AI is used more than from the tools themselves.

First, most use is individual. An employee writes a message faster, a manager summarises a report. That saves scattered minutes that never show up in the accounts. The two hours a week in the NBER paper describes this kind of use.

Second, nobody measured the starting point. If you did not know your reply time before switching on an AI assistant, you cannot know what changed afterwards. In review meetings the verdict stays an impression.

Third, AI is added to a process that is broken somewhere else. An assistant that answers enquiries in seconds and then hands the patient to a member of staff who only replies in the afternoon speeds up one part and leaves the loss where it was.

What PwC's 12% do answers those three points: they built AI into whole processes, such as demand generation and decision-making, and measured its effect with the business's own numbers.

What does this mean for a medical facility?

A medical facility has an advantage here: its repeated processes are clear and its numbers can be counted. Enquiry, reply, booking, reminder, attendance, return visit. Each step has a number. Patients themselves now use these tools before they get in touch, as our article on patients asking AI before they book explains.

Illustrative example: a four-branch group subscribed to an AI assistant to answer WhatsApp. Nobody measured reply time before switching it on. Three months later the branch manager says replies are "faster" and the finance director says bookings are "the same". Nobody has a number that settles it. Had the group measured reply time and booking rate for two weeks beforehand, it would have known within a month whether the problem was reply speed or something else.

Our Patient leak check helps you find the step where you lose the most, before you decide where AI goes.

Where does the return start in marketing and growth?

It starts where the work is high in volume, repeated, has known answers and produces a result you can count. In a medical facility that means four places, in the order we suggest:

  1. First reply and sorting of enquiries. Questions about appointments, prices, insurance and location come in every day. An assistant that answers them at once and passes any medical question to the team moves a clear number: reply time and booking rate.
  2. Reminders and follow-up. Appointment confirmations, reminders and messages after the visit. The effect shows in attendance and returning patients.
  3. Ad versions. Ad platforms now choose between many versions, and AI makes writing alternatives and resizing faster, with a medical review of every version. More in our article on AI in healthcare marketing.
  4. Reporting. Pulling numbers from the platforms and the booking system into a weekly report takes someone hours. Those hours can be saved, and the saving can be measured.

Wider administrative work, such as doctor scheduling and insurance claims, comes later because it needs deeper integration with your systems. Integration & Automation connects booking, reminders and reports, and AI Solutions builds assistants within medical limits.

How should a CFO judge an AI project?

The same way as any other investment: full cost, a number before, a number after and a rule for stopping.

In a Gartner survey of CFOs published on 20 July 2026, 45% said their AI investments lean towards productivity and 20% towards decision quality. In another Gartner survey published in 2026, 63% of finance leaders said AI had been slower than expected in 2025. Realistic expectations are part of the judgement.

Five questions a CFO can ask before approving:

  1. Which process will the project change, exactly?
  2. What is its number today, who measured it and over what period?
  3. What number makes the project a success after 90 days?
  4. What is the full cost, including subscriptions, set-up and staff time spent on review?
  5. At what number do we stop if the target is missed?

Our article on healthcare marketing ROI in a CFO's terms explains a payback calculation that applies here too.

What stays with the doctor?

Clinical decisions stay with the doctor: assessing symptoms, diagnosis and treatment advice. In Saudi Arabia, the SFDA guidance on AI and machine learning based medical devices (January 2023) requires marketing authorisation for software with a medical purpose such as diagnosis or treatment.

What a doctor can gain from AI on the growth side: fewer hours drafting, a clearer view of patients' recurring questions, and enquiries that arrive already sorted. The doctor still approves every piece of medical content published under their name. Our article on AI in the clinic sets out those limits in detail.

What are the rules in Saudi Arabia and Egypt?

The rules a leader needs to know before any project:

  • Saudi Arabia. SDAIA's AI ethics principles, including privacy and security, reliability and safety, and responsibility and accountability. Health data is sensitive data under the Personal Data Protection Law. Draft amendments to its implementing regulations, open for consultation until 5 November 2026, would tighten proof of consent before data is used for marketing. They are a draft and have not been adopted.
  • Egypt. The draft rules on doctors' media appearances, whose principles the Medical Syndicate's general assembly approved on 23 October 2026, ban false AI-made content such as claimed conference attendance, invented titles, impersonation or unreal results, according to Vetogate on 22 October 2026. The final text is still being drafted. The compliance period under the Personal Data Protection Law ended on 31 October 2026, as our article on data protection in Egypt explains.

Does AI shrink the marketing team or change it?

The available figures point to change more than shrinkage. The NBER paper found essentially no net effect on employment, and a lot of movement between roles.

In a medical facility's marketing team, what we see changing is how time is spent. Fewer hours drafting, resizing and collecting numbers. More hours on medical review, reading data, talking to doctors about patients' questions and preparing shoot days. The people who decide what gets made and review it become more important, and a team that only executes sees its role shrink.

My own view is that the most important skill a marketing team needs in 2027 is reading the booking number and tying it to what was published. That is human work.

A 90-day plan with three numbers

Pick one process, three numbers and a fixed period.

Days 1 to 15: measure. Choose the process, for example replies to WhatsApp enquiries. Measure for two weeks:

  1. Median time to first reply during working hours.
  2. Share of enquiries that end in a booking.
  3. Staff hours spent on the process each week.

Days 16 to 75: run. Switch the tool on for that process, with approved answers and immediate hand-over of any medical question to the team. Review a sample of conversations every week and log every error.

Days 76 to 90: decide. Compare the three numbers with the starting point. The decision is one of three: extend to a second process, fix what did not work and extend the trial by a month, or stop.

Time needed from management: a short weekly slot for a manager to follow the numbers, and one session at the start for a doctor to approve the answers.

In practice

The surveys say the return goes to those who tie AI to a whole process and a measured number. Start with one process in replies or reminders, and measure before you switch anything on. If you want someone to build the assistant and measure its effect with you, that is what AI Solutions is for. The wider frame for growth work is in our article on marketing company vs growth company.

Common questions

What share of CEOs see a return from AI in 2026?

In PwC's 29th Global CEO Survey, January 2026, 12% of 4,454 chief executives said AI had both raised revenue and cut costs over 12 months, and 56% saw neither. 30% saw higher revenue and 26% saw lower costs. The survey found that those with both gains were two to three times more likely to have built AI deeply into products, demand generation and decisions.

Why do most companies see no return from AI?

The figures point to three recurring causes: scattered individual use that leaves processes unchanged, no measurement of the starting point, and AI added to a process that is broken somewhere else. Companies that see a return built AI into whole processes, such as demand generation and decision-making, and measured it with their own business numbers.

Where should a clinic start using AI to see a return?

A clinic usually starts with the first reply to enquiries and sorting them, then appointment reminders and follow-up, because these tasks are frequent, repeated and have known answers. Their effect shows in clear numbers: reply time, booking rate and attendance rate. Assessing symptoms and giving treatment advice stay with the doctor.

How does a CFO measure the return on an AI project in a medical facility?

The CFO picks one process and measures its number for two weeks before launch, such as reply time, booking rate or staff hours. The full cost is then worked out, including subscriptions and review time, the number is compared after 90 days, and the project is extended, fixed or stopped according to a rule agreed in advance.

Can AI be used to write medical content in Egypt?

AI can be used for drafts, with medical review before publication. The draft rules on doctors' media appearances, whose principles the Medical Syndicate approved on 23 October 2026, ban false AI-made content such as invented titles and unreal results. The final text is still being drafted, so check its status before planning content.

Sources

More articles