A small accountancy firm owner asked me recently whether she should “go AI-native”. She had seen the phrase in a trade newsletter and assumed it meant buying a specific suite of tools. The phrase actually points to something more fundamental: a decision about how your firm is designed to operate, before any tool enters the picture. The firms doing well at AI adoption redesigned their workflows first. The product decisions came second.
What is AI-native thinking?
AI-native thinking means designing how your business runs with the assumption that AI support will be available, rather than asking whether to add a tool after the fact. Tom Davenport at Babson College describes AI-native firms as those that bake AI into core processes and decision-making. The key question shifts from “shall we try a chatbot?” to “what can AI handle here, and who in the team supervises it?”
At the AI Summit London in 2024, practitioners described four habits that characterise AI-native firms: designing workflows that assume AI support is in place; investing in data that is clean enough for AI to use; changing what staff are recognised and rewarded for; and making explicit decisions about what to stop doing to create space for AI-enabled change. The last one tends to be the hardest for owners. Adding AI to an already full plate typically means none of it takes hold.
The practical difference for a small services firm comes down to a single contrast. “We have a ChatGPT login for the team” is a tool. “Our proposal process starts with an AI draft that a senior person reviews and signs off before it goes to the client” is a redesigned workflow. The first is a bolt-on. The second is AI-native thinking applied to one process.
Why does this matter for your business?
Early AI adopters in the UK are innovating at measurably higher rates than those who wait. A rapid evidence review by the Department for Business and Trade found that first movers in machine learning saw a significantly more positive impact on innovation than second movers. For a small services firm, that gap compounds in proposals written faster, clients served more consistently, and time freed from routine administration.
The current picture shows many UK firms are still at the beginning of this. An ICS.AI survey of 2,000 UK employees in 2024 found that only 10 per cent of organisations qualified as AI-native, with AI deployed across the business rather than in isolated pilots. A further 36 per cent had implemented AI in specific teams only, and 20 per cent had not implemented it at all.
The deeper competitive argument is about being first to change how decisions get made day to day. Experimentation on its own moves nothing. An owner who has reorganised a reporting workflow so it takes twenty minutes instead of two hours has not just saved time. She has created capacity for the work that pays better and draws more on judgement, and less on administration. That is a structural change in what the firm can do, and it compounds across every month it runs.
Where will you actually meet it?
AI-native thinking shows up in the recurring, information-heavy work your team does every week: proposals, client reporting, inbox triage, meeting notes. The common thread is information work where AI handles the first pass and a human reviews the result before it goes anywhere. For firms of five to fifty people, these tasks absorb a disproportionate amount of time, and they are also the clearest starting point.
ICS.AI’s work with Derby City Council illustrates the pattern at scale. An AI assistant handled over 40,000 contacts in its first three months, resolving around 75 per cent without escalation to a human agent. The humans remained in the loop for edge cases, complaints, and anything requiring discretion. For a small services firm, a well-designed inbox filter, a first-pass proposal tool, or an AI assistant over your standard operating procedures already qualifies. The council-scale deployment can come later, if it ever makes sense at all.
Law firms offer another illustration. Allen and Overy’s pilot of the AI legal assistant Harvey allowed lawyers to produce first-cut documents and research faster, with the firm explicit that human lawyers reviewed and remained responsible for every client-facing output. The workflow changed; the accountability didn’t. That distinction is central to AI-native thinking. The question is always who reviews and signs off, not whether AI was involved.
When to ask the AI-native question, and when to hold off
The AI-native framing is most useful when your team does the same information-heavy work repeatedly and the bottleneck is time spent on routine drafting rather than the thinking itself. It becomes a liability when your data foundations are weak, when basic compliance gaps exist, or when the core bottleneck in your service is physical skill rather than information handling. Getting the diagnosis right first saves expensive backtracking.
Regulation is a practical consideration here, not a theoretical one. The UK Information Commissioner’s Office is explicit that AI processing of personal data must comply with UK GDPR, including data minimisation and lawful basis. Where AI processing could create high risk for individuals, such as automated scoring or filtering, the ICO expects a Data Protection Impact Assessment before you deploy. The National Cyber Security Centre recommends treating AI tools as part of your existing cyber-risk posture, with the same access controls and supply-chain scrutiny you would apply to any other third-party software.
Cost is worth confronting directly. Microsoft 365 Copilot launched at around $30 per user per month. For a twenty-person firm, that is a material line on the profit and loss account, and without process redesign, the return rarely materialises. The firms that see genuine returns are those that pair the tool with a changed workflow, not those that add it to an unchanged one. The EU AI Act adds a further consideration: SMEs serving EU clients may be treated as deployers of high-risk systems and face risk management and human oversight obligations if they use AI in areas such as recruitment or credit assessment.
What this means for your team and how decisions get made
The staff-readiness picture in the UK is more optimistic than many owners assume. An ICS.AI survey of 2,000 UK employees found 62 per cent were excited about AI’s potential at work. The same survey found more than 80 per cent wanted humans to remain involved in AI-backed decisions, with 46 per cent saying humans should give final approval on AI suggestions, and 37 per cent preferring human-led decisions supported by AI data.
That means your team is probably open to this, but they want clarity on accountability. They want to know who is responsible when AI gets something wrong. A short written AI use statement, covering what AI can be used for and where human sign-off is required, addresses that before it becomes friction. Firms that skip this step tend to see uneven adoption, where some staff use AI heavily and others avoid it entirely because the ground rules are unclear.
One practical step worth taking from the start: plan for vendor continuity. A Red Hat survey of 500 UK and EU IT decision-makers in 2024 found that over two-thirds of UK companies already had an exit strategy in place if their primary AI provider restricted access. Keeping your key prompts, workflow documentation, and output records in a portable format costs almost nothing now and is considerably harder to retrofit after two years of dependency on a single provider.
If you want to talk through which workflows in your firm fit the AI-native pattern and which don’t, Book a conversation.
