Type to search every page. Use the arrow keys to move through results.
Industries
Public sector
Leaders in government, healthcare and education answer to legislatures, patients, students and families for how public money and public trust are used. Published research shows AI in use in all three, most often on planning, writing and documentation work, with the rules and skills around it still being built.
The strategies, policies and skills that should surround that use are still being put in place. Of 87 UK central government bodies surveyed by the National Audit Office in autumn 2023, 21% had an AI strategy. In the same survey, 70% agreed that difficulty recruiting or retaining staff with AI skills was a barrier. Around one-fifth of schools in England had a policy on the safe and appropriate use of AI, according to the Department for Education.
Leaders in the public sector answer for these choices to legislatures, boards, patients, students and families, often in public. The questions that follow are which uses to back, who is accountable for each and how the results will be shown. We help you set the strategy, govern the tools you keep, train your people and report on the results in terms an auditor, a board or a family can follow.
The figures on this page are third-party research. Each links to the organization that published it, and none describes our own work.
In the data
Statistics Canada asked businesses across industries about their plans for AI. Health care and social assistance, at 33.8%, is one of four industries above the all-industry figure of 25.2%. The table has no row for public administration or educational services.
Share of Canadian businesses planning to use AI over the next 12 months, by industry% of businesses in each industry, third quarter of 2026 (9,601 responses)
Professional, scientific and technical services54.8%
Information and cultural industries48.0%
Finance and insurance35.5%
Health care and social assistance33.8%
All industries25.2%
Arts, entertainment and recreation24.5%
Manufacturing24.4%
Real estate and rental and leasing23.6%
Administrative and support services19.6%
Construction18.3%
Accommodation and food services17.4%
Wholesale trade16.8%
Other services (except public administration)16.7%
Retail trade16.0%
Mining, quarrying, and oil and gas extraction14.8%
Agriculture, forestry, fishing and hunting13.2%
Transportation and warehousing8.0%
These figures are plans, and the survey table does not measure use. The health care row covers health care and social assistance as a whole. The survey’s table covers 16 industries, all shown here.
Leaders in government answer to legislatures, auditors and citizens for the services they deliver, the money they spend and the systems that both depend on. The evidence here comes from national governments and their audit offices. None of it describes provincial, state or municipal bodies.
Just over a third of the UK central government bodies surveyed had deployed AI, and about one in five had an AI strategy.
Where UK central government bodies stood on AI% of 87 central government bodies surveyed, autumn 2023
37%had deployed AI
70%were piloting or planning AI
21%had an AI strategy
The three shares are reported separately and are not parts of one whole. The survey covered departments and arm’s-length bodies above about £83 million in operating spend. A further 61% had plans to develop an AI strategy.
We work with your service leaders to choose the AI uses that address your most pressing service problems, and plan how each will be delivered and measured.
Leaders answer to auditors and legislatures for payments made and money lost. AI can speed up the identification of suspect payments, so that reviewers reach them sooner.
The US Department of the Treasury says that using machine learning to expedite the identification of Treasury check fraud resulted in US$1 billion in recovery in fiscal year 2024, part of more than US$4 billion in fraud and improper payments prevented and recovered. It is the department’s own statement about its work.
Source: US Department of the Treasury
Our part
We build automations and AI agents that gather the evidence behind a flagged payment for your reviewers, who make each decision.
New tools depend on the systems and data a department already has. In the National Audit Office’s survey, 62% of UK central government bodies agreed that a lack of available or good-quality data was a barrier to AI.
Saying where a government uses AI, and who answers for it, is part of showing how the work is done. Canada’s federal government has published a register of AI uses: over 400 systems where AI is being explored, developed, implemented or deployed, with contributions from 42 federal institutions.
Our part
We set up governance for AI use: an owner for each system, a stated purpose and a record you can show an auditor or the public.
A ministry needs people who can choose, run and question AI tools. Canada’s federal AI strategy says there is a 30% vacancy rate for digital roles in the Government of Canada, which it says threatens delivery.
In the National Audit Office’s survey, 70% of UK central government bodies agreed that difficulty recruiting or retaining staff with AI skills was a barrier.
Source: UK National Audit Office
Our part
We train your staff to use and oversee AI tools, and help you hire the skills you lack.
Funding. Of the 87 UK central government bodies surveyed, 70% agreed that lack of funding for development or implementation was a barrier to AI. Source: UK National Audit Office
Legal and accuracy risk. In the same survey, 67% agreed that lack of clarity on legal liability for AI use cases was a barrier, and 57% agreed that the risk of inaccurate or unreliable outputs was. Source: UK National Audit Office
Privacy and public trust. Of the same bodies, 56% agreed that the risk of privacy, data protection or cyber security breaches was a barrier, and 43% agreed that a lack of public trust in AI was. Source: UK National Audit Office
Existing policy can be an obstacle. Officials at 10 of the 12 federal agencies GAO selected said that existing federal policy could present obstacles to adopting generative AI. Source: US Government Accountability Office
Most use is early. Of the 282 generative AI use cases GAO counted for 2024, 84 were at the first stage, where need and purpose are documented, and 75 were in acquisition or development. Source: US Government Accountability Office
The questions we start with
Which of the services your organization is judged on could an AI tool improve, and who is accountable for each?
Could you show an auditor which AI tools are in use, what each is for and who owns it?
Which of the systems behind your main services are too old for an AI tool to rely on?
How many of the people who would run an AI tool are already on your staff, and how many would you need to hire?
Which payments or claims are reviewed by hand today, and how much staff time does that take?
Healthcare systems and public health authorities
Leaders in healthcare answer for access, quality and safety, and for the budget behind them. The Canadian Institute for Health Information forecast that total health spending in Canada would reach $399 billion in 2025, or $9,626 per Canadian, equal to 12.7% of GDP.
AI is already in use in clinics and hospitals. In the 2025 Commonwealth Fund survey, 46% of Canadian family doctors reported that their office used AI. In the US, 71% of non-federal acute care hospitals used predictive AI integrated with their electronic health record in 2024, up from 66% in 2023.
Among US hospitals that use predictive AI, use for billing and for scheduling grew between 2023 and 2024.
What US hospitals using predictive AI use it for% of hospitals using predictive AI
20232024
Simplifying billing2023: 36%2024: 61%
Facilitating scheduling2023: 51%2024: 67%
Each share counts only the hospitals that use predictive AI in that year. The data brief reports use and does not report what it saved.
Leaders answer for how scarce clinician time is spent, and documentation takes a share of it. In the Commonwealth Fund survey, 68% of the Canadian family doctors’ offices that used AI used it for notetaking, documentation or scribing, 11% for patient triage and 8% for care planning.
Canada Health Infoway reports that in its national AI Scribe Program nearly 70% of clinicians experienced reduced administrative burden, and that clinicians saved an hour or more a week on average. Half of them redirected the time to other administrative tasks and 20% to more appointments. The results are reported by the clinicians themselves.
Source: Canada Health Infoway
Our part
We study in person how your clinicians document care and which systems they use, then build a Consier AI solution around that.
Leaders answer for patient safety, and for who is responsible when a tool is wrong. In the US, hospitals that already use predictive AI mostly evaluate it and monitor it once it is in use.
In 2024, 82% of US hospitals using predictive AI evaluated models for accuracy and 79% monitored them after introduction.
Source: ASTP/ONC
Our part
We set up governance for clinical AI: an owner for each tool, a stated purpose and a record of how it is checked.
Public health agencies answer for the health of whole populations and report on it to legislatures and the public. A 2026 profile asked state and territorial agencies in the US how they use AI.
Among 44 US state and territorial health agencies answering, 30% used AI for administrative and operational efficiency and 30% for content and report creation, while 32% were not using AI. The sample is small.
Source: Association of State and Territorial Health Officials
Our part
We build the reports and dashboards your analysts and leaders use to track the health of the population and the performance of programs.
Bias is checked less often than accuracy. Of the US hospitals using predictive AI in 2024, 74% evaluated models for bias, and 66% had a specific AI committee or task force. Source: ASTP/ONC
Data held in separate systems. The OECD’s 2025 review of health systems says that fragmented data systems limit the potential of analytics and AI. Source: OECD
Results are reported by those who use the tools. The figures Canada Health Infoway publishes for its AI scribe program are clinicians’ own reports. Source: Canada Health Infoway
The questions we start with
How much of your clinicians’ working week goes to documentation, and which of your sites have tried tools for it?
Which clinical AI tools are already in use in your system, and who answers for each of them?
Do your smaller and rural sites have access to the tools your largest sites use?
Could you show a regulator or your board how each clinical tool is evaluated and monitored once it is in use?
Where is patient data held, and could one tool read it across your hospitals, clinics and public health teams?
Education
Leaders in education answer to students, families and funders for what is learned, how it is assessed and how staff time is spent. AI now bears on all three.
Leaders answer for whether the work students hand in is their own and whether grades mean what they say. In England’s 2025 survey, 77% of secondary school leaders whose pupils could access generative AI had experienced or been made aware of issues as a result. Plagiarism (67%) and exposure to misinformation (53%) were the most common. The report counts leaders who met an issue and gives no measure of how much plagiarism occurs.
Our part
We help you set what each AI tool may be used for at each level of study and how work is assessed, and give each approved tool an owner.
Teachers answer for planning, feedback and administration, and leaders decide which tools are worth buying. Where schools in England had invested in AI-related tools, the most common purchase was added AI capability in software they already used.
In the same survey, 51% of IT leads said their school had not invested in any of the listed AI-related tools since January 2023. A further 41% had invested in at least one: 34% in increased-capability licences for existing software, 10% in licences for generic AI tools and 7% in education-specific generative AI tools.
Source: Department for Education (England)
Our part
We build automations and AI agents for routine administrative and planning work, with your staff reviewing the results.
Leaders have to say what is allowed in each course, what is assessed and how data is protected.
In England in 2025, around one-fifth of schools had a policy on the safe and appropriate use of AI: 34% of secondary schools and 20% of primary schools. Interviewees more often described informal guidance than a formal policy.
Source: Department for Education (England)
Our part
We develop your AI strategy and the policy that follows from it, covering which uses to support, which to restrict and how each is reviewed.
Of the UK higher education students surveyed in 2024/25, 24% said they had received training or support on using AI appropriately and effectively.
Source: Jisc
Our part
We train your teachers and administrators to use and oversee AI tools, and prepare some of your own staff to support their colleagues.
Trust in the tools. Interviewees in England’s survey cited the limits of current tools, distrust of their outputs, difficulty in phrasing requests to them, attitudes to AI use and cost. Source: Department for Education (England)
Provision is still limited. Of the UK higher education students surveyed in 2024/25, 23% said their institution provided AI systems, chatbots or assistants, up from 11% in the previous year’s survey. Source: Jisc
Rules are still being written. In UNESCO’s 2025 survey, 19% of respondents said their institution already had a formal AI policy and 42% said frameworks were under development. One in four said their university had already encountered ethical issues linked to AI. The survey had 400 responses from UNESCO Chairs and UNITWIN Networks at higher education institutions in 90 countries. Source: UNESCO
The questions we start with
Which AI tools do your students and staff use today, and which of them has your institution approved?
How does each course say what is permitted, and how are staff expected to assess work in light of it?
How many of your teachers have had training in AI, and what would each of them need next?
Which administrative tasks take the most staff hours, and which of those suit an automated tool?
Who reviews the data protection of each AI tool before it reaches a classroom?
How we can help
These are problems in the sectors above that AI is being used on today. For each, we set out what published evidence shows and what we would do.
Canada’s federal government has set out to reduce what it spends on its own operations while keeping the services Canadians depend on. Budget 2025’s Comprehensive Expenditure Review is to save $9 billion in 2026–27, $10 billion in 2027–28 and $13 billion in 2028–29.
The same budget expects the public service to fall from almost 368,000 people in 2023–24 to roughly 330,000 by the end of 2028–29. More than half of the budget’s total savings come from modernizing government operations. Several departments plan to use AI and automation for routine and back-office work.
What AI is doing about it today
The UK government tested an AI assistant built into the office software its staff already used. In the Government Digital Service’s trial, 20,000 employees used it from September to December 2024. Participants reported saving an average of 26 minutes a day. These were their own estimates, and the trial could not show how the time saved was spent.
One department measured more closely. The Department for Business and Trade’s evaluation of its pilot, with 1,000 licences, found small time savings on most tasks and the largest on written work. Some tasks, such as scheduling, took longer. It found no evidence that the time saved had improved productivity, and it identified training as a significant factor in how satisfied staff were.
Yearly savings planned under Canada’s Comprehensive Expenditure ReviewSavings in each fiscal year, $ billion
Work alongside your teams to measure where staff hours go on routine work such as correspondence, briefing notes, case summaries and internal service requests.
Choose the few uses that matter most to your savings plan, and compare the tools open to you, including those you already license, with no preference for any supplier.
Run each pilot against a comparison group, measuring time, quality and errors directly as well as asking staff for their estimates.
Train staff in the approved tools and in checking what they produce, with written rules for what each tool may be used for.
Report the measured results in hours, cost and service levels, in a form your executive committee can act on and an auditor can follow.
What you would have
You would have a measured account of where staff hours go and a short list of uses tied to your savings plan. You would also have pilot results measured against a comparison group, trained staff with written rules for each tool and a report your executives can act on.
A missed appointment leaves a clinic slot empty while other patients wait. NHS England reported in March 2024 that eight million of the 124.5 million outpatient appointments in England in the previous year were not attended, or 6.4%. It estimated the yearly cost to the NHS at £1.2 billion.
The rate is highest in some specialties, as the graphic shows. NHS England presented its wider use of AI against missed appointments as a way to improve waiting times.
What AI is doing about it today
Mid and South Essex NHS Foundation Trust piloted software that predicts which patients are likely to miss an appointment and offers back-up bookings. Over six months, NHS England reports, non-attendance fell by 30%, 377 missed appointments were prevented and 1,910 more patients were seen. In March 2024 NHS England said the software would be rolled out to ten more trusts.
Sheffield Children’s NHS Foundation Trust used an AI tool to send an extra reminder, with an offer of support, to patients judged at least 50% likely to miss their appointment. On a benchmark non-attendance rate of 19.27% it expected 8,581 missed appointments over the first 12 months, and just under 6,500 were recorded. These are the trusts’ own pilot results, as NHS England reports them.
Outpatient appointments not attended in England, by specialty% of appointments not attended, in the year NHS England reported in March 2024
Physiotherapy11%
Cardiology8.9%
Ophthalmology8.8%
Trauma and orthopaedics7.9%
All outpatient appointments6.4%
The four specialties NHS England named as highest, with the rate for all outpatient appointments: eight million of 124.5 million.
Sit with your booking teams and clinic managers, and measure missed appointments by specialty, site and patient group, with the clinic time each one costs.
Find what lies behind the misses in your own records, such as reminder timing, appointment times and travel, then compare the tools open to you with no preference for any supplier.
Build automations that send reminders, offer rebooking and fill freed slots, with your booking staff deciding who is offered what.
Give the tool an owner and a stated purpose, review its use of patient data and check that its predictions do not disadvantage any group of patients.
Build a monthly report that shows missed appointments, refilled slots and waiting times by specialty and site.
What you would have
You would have a measured baseline of missed appointments and their causes, and a documented choice of tool. You would also have automations your booking staff use every day, a record of how the tool is governed and checked, and a monthly report by specialty.
In England’s 2025 survey of teachers and leaders in state schools, full-time teachers reported working an average of 50.1 hours in their most recent term-time week. Only 26% of teachers and leaders agreed that the amount of work they had was acceptable.
In the same survey, 29% were considering leaving the state school sector within 12 months for reasons other than retirement. Of those, 89% cited a high volume of work, as many as cited stress or poor wellbeing.
What AI is doing about it today
The Education Endowment Foundation published a trial, independently evaluated, with 259 science teachers in 68 secondary schools in England. Teachers given a freely available generative AI tool and a written guide spent 56.2 minutes a week preparing lessons and resources. The comparison group spent 81.5 minutes, so the tool group spent 31% less time.
An expert panel that did not know which resources used the tool found that quality did not appear to be affected. Teachers tended to use it for one or two tasks, most often quizzes and activity ideas, and had a five-week learning period before their planning time was recorded. In the tool group, the share who felt they spent too much time on preparation fell from 49% to 26%.
Weekly time spent preparing lessons and resources in the trialMinutes a week, Year 7 and 8 science teachers in England
Comparison group, no generative AI81.5 minutes
Generative AI tool with a guide56.2 minutes
A trial of 259 teachers in 68 schools, evaluated independently. The difference is 25.3 minutes a week, or 31%.
Spend time with teachers and heads of department to measure where their hours go outside the classroom, from planning and feedback to administration.
Set out which AI tools staff may use and for what, what student information must never be entered and who owns each approved tool.
Train teachers in the approved tools, with time to practise before results are measured, and prepare some of your own staff to support colleagues.
Build automations for routine administrative work, such as letters to families and attendance follow-up, with staff reviewing everything before it is sent.
Measure preparation time and the quality of materials before and after, and report the results to your board or governors.
What you would have
You would have a measured account of where teachers’ time goes and written rules for the tools you approve. You would also have trained staff with colleagues who can support them, automations for routine administrative work and a before-and-after measure of time and quality for your board.
The results described here were published by the organizations named. These are examples of how we would approach the work, and none describes an engagement of our own.
The UK’s national audit office surveyed central government bodies on how far they had deployed AI, whether they had a strategy and what held them back.
A briefing is a confidential conversation with the executive accountable for a technology decision. We set out how we would approach the question and what evidence it would require. Scope and fees are agreed in writing before any work begins.
