Before You Build an AI Strategy, Find the Workflow Worth Automating

AI adoption is often talked about as if every business is already moving at speed and anyone hesitating is falling behind. The latest GOV.UK AI Adoption Research gives a more useful picture. Across 3,500 UK private sector businesses, only 16% said they currently use at least one AI technology. Another 5% said they plan to adopt AI in future, while 80% neither use AI nor have plans to do so.

That does not mean UK businesses are ignoring technology. It means the practical route into AI is still unclear for many of them. More than half of businesses surveyed, 51%, said they do not see AI as relevant to their organisation. When asked what prevents or has previously prevented adoption, the most common answer was not cost, regulation or ethics. It was that they had not identified a use for AI in the organisation, cited by 71% of businesses.

For SMEs, this is the real starting point. Most do not need an abstract AI strategy first. They need to find one piece of work where AI and automation could remove friction, reduce manual effort, or make a process more reliable. The strategy can grow from evidence. The first useful workflow is what creates that evidence.

The research also shows why a tool-first approach can disappoint. Among businesses already using AI, 85% are using natural language processing and text generation. That is not surprising. These tools are accessible, easy to test and useful for drafting, summarising and searching. But the same report found that agentic AI adoption is only 7%, and that businesses are most likely to face significant barriers when implementing agentic AI, at 32%.

In plain terms, many organisations have reached the easy part of AI: giving people tools that help with individual tasks. Far fewer have reached the harder and more valuable part: embedding AI into a dependable business workflow. That distinction matters because a business does not become meaningfully more efficient just because someone can draft an email faster. It becomes more efficient when the surrounding process changes.

Take enquiry handling. A member of staff using AI to draft replies might save a few minutes at a time. A workflow that captures the enquiry, checks whether key information is missing, classifies the request, drafts the response, updates the CRM and flags the right follow-up can change the way the business operates. One is individual productivity. The other is operational improvement.

The GOV.UK research points to that gap. Three quarters of AI-using businesses reported improved workforce productivity, and 57% said AI had helped develop new or improved processes or operations. But 77% had not yet seen a change in revenue since adopting AI, while only 12% reported an increase. Productivity gains are useful, but they do not automatically become commercial gains unless they are attached to a process that affects speed, capacity, quality, conversion or service delivery.

That is why SMEs should start with workflow selection, not technology selection. The best first AI project is usually hiding in repeated work that already frustrates the business. It might be sorting inbound enquiries, preparing quotes, chasing missing information, updating records after calls, compiling reports, checking documents, triaging support requests, reconciling spreadsheet data, or producing routine customer updates.

The right workflow has a few signs. It happens often enough to matter. It uses information that can be accessed consistently. It has a clear before-and-after state. It includes repetitive judgement, preparation or routing work, but still has obvious points where a human should approve the output. It creates delays, errors or admin drag that people already complain about.

This approach also helps avoid overbuying. The research found that the most common areas where businesses are using or planning to use AI are marketing and administration, both at 72%, followed by IT at 64%. Those are sensible starting points because they contain a lot of repeatable work. But the useful question is not whether AI can help marketing, admin or IT in general. It is which exact workflow inside those functions is worth fixing first.

For example, a marketing team might not need a general AI content tool. It may need a process that turns customer questions, sales notes and project examples into useful draft content for review. An admin team might not need another chatbot. It may need an automation that checks forms, flags missing data and updates the right system. An IT team might not need a broad AI platform. It may need a support triage workflow that categorises requests and routes them properly.

This is a healthier way to think about readiness as well. GOV.UK found that 54% of organisations already using AI feel ready to scale their use, but among those only planning to adopt AI, readiness drops to 34%. That gap is understandable. It is hard to feel ready for "AI" as a broad category. It is much easier to get ready for one defined workflow with clear inputs, outputs, risks and ownership.

Good AI adoption should therefore begin with a short operational audit. Where is time being lost? Which tasks depend on someone remembering to move information between systems? Which customer or internal processes slow down because data arrives in the wrong format? Which reports, updates or checks happen repeatedly? Which work is valuable but boring enough that it often gets delayed?

Once that workflow is identified, the implementation becomes more grounded. The business can map the current process, define what success looks like, decide where AI helps, set the boundaries, test with real examples and measure the result. That is much more practical than trying to write a broad AI roadmap before the company has learned where AI actually fits.

This also keeps the risk sensible. A small workflow can have human review, logs, fallback behaviour and clear ownership. It can be improved over time. If it works, it becomes the pattern for the next workflow. If it does not, the business has learned something without turning AI adoption into a large, expensive programme.

For SMEs, the opportunity is not to copy how large organisations talk about AI. It is to use their own advantage: shorter decision paths, clearer operational pain and faster implementation. A smaller business can pick one workflow, build properly, learn quickly and move on to the next useful improvement.

The businesses that get value from AI will not necessarily be the ones that bought the most tools first. They will be the ones that turned the right tools into reliable work. GOV.UK's research suggests many organisations are still waiting for a clear use case. For SMEs, that use case is usually already there. It is the repeatable workflow that everyone knows is slower than it should be.

Birdcage Tech helps SMEs turn AI ideas into practical workflows that reduce manual work, improve consistency and fit the way the business actually operates. If you want to move beyond AI curiosity, start by finding the workflow worth automating.

Source note: this article references GOV.UK's AI Adoption Research, updated 13 February 2026, published by the Department for Science, Innovation and Technology.