The Office for National Statistics reports that AI use among UK businesses with ten or more employees has risen from around 12 per cent in late 2023 to around 35 per cent in mid 2026. Over the same period the average number of AI technologies in use per business moved from 1.4 to 1.6. RICS’s 2026 survey of more than 3,100 chartered surveyors shows the same shape in the built environment: two thirds of construction professionals and three quarters of those in commercial property now use AI, but most of that use is at the pilot stage and embedded use remains rare.
Training has not kept pace. According to SAP and Oxford Economics, four in ten UK businesses have given their staff comprehensive AI training. The Access Group found that 70 per cent of UK employees are experimenting with AI at work without any formal training. Many use whatever they can find. A Microsoft study of 2,003 UK workers found that 71 per cent had used unapproved consumer AI tools at work and 51 per cent do so every week. MIT’s 2025 review of enterprise generative AI found that around 95 per cent of pilots produced no measurable effect on profit and loss, and put this down to a learning gap in organisations rather than to the models.
For surveyors there is now a regulatory requirement as well. Since 9 March 2026, RICS’s professional standard on the responsible use of AI in surveying practice has required members and regulated firms to maintain sufficient knowledge of the tools they use, keep a written register of AI systems, apply and document professional judgement over outputs, and tell clients in advance when AI will be used. RICS’s own survey found that a quarter of construction respondents did not know the standard existed. A firm cannot meet those requirements unless its people understand the tools they are using.
What we see in organisations where AI has become routine is a small number of repeated activities, kept going for a year or more, rather than a launch campaign. We call them rituals. They keep AI visible, make learning ordinary and give leaders a regular read on what is working. The list below sets out the rituals we have deployed with clients across sectors and which have worked. None needs a significant budget and most can run on the collaboration and intranet tools an organisation already has. They are grouped into four themes. No organisation needs all of them; choose the ones that fit your structure, your stage on the journey and the initiatives already under way.
1. Learn: building capability
Role-specific training, in two tiers. Build the training around people’s own work: for a surveyor, analysing a lease, checking a specification against a schedule of condition or summarising a planning report, using the firm’s own data governance rules. Run two tiers. A beginner course covers what AI is, its limitations and risks, how to write a good prompt and how to check an output. An advanced course covers building and orchestrating agents, connecting them to data sources and managing them once live. Allow two to three hours per tier, delivered in person or live online, with a short refresher each quarter as the tools change. For regulated firms, the beginner tier goes a long way towards the RICS standard’s knowledge requirement.
Themed Lunch and Learn sessions and AI clinics. A weekly or fortnightly 45-minute slot with a published theme, a 30-minute live demonstration and then an open clinic where people bring their own problems. Publish the next three to five themes in advance so people can pick the sessions relevant to them. Frame each theme as a task: “How do I extract the key terms from this lease?” draws more people than “Introduction to document summarisation”. Record the demonstration and post it afterwards. Attendance holds up when people know they will leave with something they can use that afternoon.
A weekly “How do I do this with AI” guide. Ask teams what they wish they could do with AI and how they do it today, then run a short poll so people rank the ideas. Each week, write a one-page guide to the top-ranked task: the steps to follow, which tool and data sources to use, and how to check the result. File every guide in a searchable catalogue on the intranet so the collection builds into a reference.
A library of short screen recordings. A day spent recording three to five minute clips of common AI tasks, narrated and subtitled, produces a resource people return to. Catalogue the clips on the intranet or help site next to the written guides. With one client, individual clips were viewed by thousands of colleagues, mainly because they showed exactly what to click and what to type. Refresh the library each quarter as interfaces change.
An organisation-specific glossary. Write it in the organisation’s own language: what “grounding” means when the data is your lease database, what the data governance rules allow and prohibit in practice, and which tools and use cases each term relates to. Keep it to a single page, link to it from every other resource and review it every six months.
Manager briefings and new-joiner induction. Line managers decide, in practice, whether their teams experiment, so give them a one-hour briefing on what to expect, what to encourage and where the guardrails are. Add a 20-minute AI module to the induction pack so every new joiner arrives knowing the policy, the approved tools and where the resources live.
2. Share: making AI use visible
Show and Tell sessions. Set aside the last hour or two of the week, fortnightly or monthly, for colleagues to present what they have done with AI since the last session, large or small. Choose and brief presenters in advance. Ask each to cover how they set it up, what went wrong, what the risks were, what the output looked like and how it changed their work. Failures are as useful as successes. The sessions give other people ideas, surface blockers and give the programme team a running list of where people get stuck, which feeds the improvement backlog.
An “AI moment” in team meetings. A standing two-minute item at the start of regular team meetings where one person describes something they did with AI that week. It costs nothing and reaches the people who never attend optional sessions. Rotate the speaker so it does not become the same enthusiast every week.
A weekly “What’s New in AI” note. A few informal paragraphs each week recommending one book, article, video or podcast, with a note on why it matters and how it applies to the organisation. Send it within an existing weekly communication or as a standalone end-of-week note. Ask a business leader to contribute their own pick once a month.
A community channel and prompt of the week. One channel on the organisation’s messaging platform where people post prompts, questions and results, seeded by the AI champions. Pin a prompt of the week every Monday. The channel keeps things moving between scheduled sessions and is the natural home for the prompt library.
A shared prompt library. Collect prompts that work, organised by role and task, with a note on which tool and data source each is designed for. Start with a spreadsheet; a simple internal app built on a workflow tool can follow once the library has proved its value. Ask each team to contribute five prompts to seed it, and review it monthly to remove the ones nobody uses. For surveying firms, organising by service line (valuation, building surveying, project management, planning) is more useful than organising by tool.
Automated audio and video summaries of key meetings. Few people will sit through a two-hour recording of a monthly reporting forum. Most meetings already produce a slide deck and a transcript, and combining the two into a 15-minute audio summary and an eight-minute video is now easy to automate. Send both to participants and to those who could not attend. Engagement with governance forums rises, and the whole organisation sees a practical use of AI every month.
A lunchtime AI competition. A light-hearted event with a small prize, run across offices: “What is wrong with this prompt?”, “Spot the risky behaviour”, “Build an agent in five minutes”. A large screen and 45 minutes is enough. Run it quarterly and tie it to the current training themes.
An agentic hackathon. A half- or full-day event where attendees, ideally the AI champions, each bring a repetitive process and build an agent to automate it with specialist support. A series of three sessions a month apart works best: the first automates a process, the second connects that automation to other processes and data sources, and the third covers running several agents together. Cover production, aftercare and maintenance in each session so the agents are still in use six months later.
3. Listen and measure
A monthly pulse poll. A three-to-five minute poll each month, sent by a senior executive, on how people are using AI, how often, what is blocking them, what they want to learn next and what help they need. Keep the core questions the same each month so trends are visible, and change one or two to probe current issues. Use the results to set the training themes, the weekly guides and the order of the tool roll-out.
A live use case dashboard. A catalogue of AI use cases, visible to everyone, showing what is planned, in progress and done. Let colleagues add their own ideas and vote on others. A simple dashboard built on a spreadsheet is enough to start. It shows people the overall direction and their place in it. For RICS-regulated firms, it can double as the written register of AI systems the professional standard now requires.
An adoption scorecard and time-saved log. Combine licence usage data from the tool vendor, attendance at sessions and a simple monthly “hours saved” self-report from champions. Review it monthly. This turns anecdotes into a business case and tells you which of these rituals are earning their place and which to drop. BCG’s 2026 AI at Work survey found that 42 per cent of regular AI users report saving at least eight hours a week; a scorecard lets an organisation check whether that holds for its own people.
4. Lead, govern and enable
A monthly executive update on the AI journey. Each month, someone senior sets out what has been achieved, what blockers were met and what is coming next, with successes and failures both covered. BCG found that only a third of employees say leadership communications about AI are clear, and that a clear strategy matters more to adoption than access to tools. Leaders describing how they personally use AI has a noticeable effect. The update is also the place to point back to every resource above.
A plain-English AI use policy. A short, readable policy covering the approved tools, who owns which data, permitted and prohibited uses and how AI use is governed. Put it where people will find it, not in a document management folder. SAP’s research found that 68 per cent of UK businesses believe their staff uses unapproved AI tools at least occasionally; a visible policy reduces that risk and gives people confidence to use the approved tools. For surveying firms, the policy should reference the RICS standard directly and set out how the firm meets its requirements on registers, dip sampling and client transparency.
A cohort of AI champions. Identify the early adopters across departments and business units, train them, give them early access to tools and allow them dedicated time (half a day a week is a reasonable start) to help others, run sessions and feed back. Define the role in writing, with a clear commitment, recognition or reward, and an explanation of how it helps their career. Pairing works well as an extension: a champion spends an hour with a more hesitant team working on one of that team’s real tasks.
Licence mapping and a route to advanced tools. Everyone should have a base licence for the organisation’s chosen assistant. Map advanced licences and specialist tools to roles rather than issuing them on request. Beyond that, publish a pre-approved catalogue of additional tools that any colleague can apply for on a three-month trial, after which they show what they have achieved to keep the licence for a further year. This keeps the tool estate under control and gives curious people a legitimate route, which removes much of the reason for shadow AI.
Choosing where to start
The pattern in organisations where AI use has become routine is four or five of these rituals running in parallel for a sustained period, with a named person responsible for keeping each one going.
Start with what is already moving. If a team is already running informal demonstrations, give them a theme calendar and record the sessions. If a prompt spreadsheet exists, publish it and ask for contributions. Then add one ritual from each of the other themes so that learning, sharing, listening and leadership are all covered. Review the scorecard after three months, drop what is not being used and add the next ritual.
The UK government has set a target of ten million workers with key AI skills by 2030. For most organisations the quickest contribution to that is a standing item in the team meeting and a clinic on Wednesday lunchtime, run every week until nobody thinks of them as an AI initiative any more.
