Deloitte surveyed 3,200 leaders about AI. Here's what it means if you run a £5m business.
84% of organisations are increasing their AI spend this year. At the same time, 37% are using AI with zero change to their underlying processes. They are buying tools and leaving them on the shelf.
If you run a business turning over £1m to £25m, you've probably felt the pressure. AI is everywhere. Every conference, every LinkedIn post, every software vendor pitch. The implication is clear: if you have not started adopting AI, you are falling behind.
But most business owners aren't interested in AI for its own sake. What they want is straightforward. Automate the repetitive work. Speed up the processes that drag. Free up time for the work that generates revenue. AI should be the means to that end.
The problem is that most founders don't know where to begin. And the guidance available is written for companies with 500 employees and a Chief AI Officer, not for someone running a 30-person business who may still approve every invoice personally.
Deloitte's "State of AI in the Enterprise 2026" report surveyed 3,235 business and IT leaders across 24 countries. The findings are aimed at large organisations. But three patterns in the data apply directly to AI adoption in small business, and in some cases, they are more pronounced at the £1m-£25m level, where there's no dedicated team to bridge the gap between buying a tool and making it work.
Each of those three patterns points to the same root cause. You can't embed AI into operations you haven't structured yet.
We've seen this first-hand. A client brought in a third-party AI provider to build use cases for their business. Before any AI work could start, the data had to be centralised. It lived in different systems, in different formats, owned by no one in particular. Processes had to be rerouted. Data owners and stewards had to be assigned and trained. A data privacy function had to be established. The AI project became an operations project before it became anything else.
That pattern repeats at every scale.
You bought the AI tools. Your team isn't using them.
When it comes to AI adoption, small business leaders assume access is the hard part. Deloitte found that workforce access to AI tools grew by 50% in a single year. Under 40% of workers had access to sanctioned AI tools in 2024. By 2025, that figure was closer to 60%.
But access didn't translate into use. Fewer than 60% of workers who had AI tools available used them in their daily workflow. That number hasn't changed from the year before. Companies gave their teams the tools and nothing happened.
Meanwhile, 37% of companies surveyed are using AI at a surface level, with little or no change to existing processes. They have plugged in a tool without changing how the work gets done. The tool sits alongside the old way of working rather than replacing any part of it.
In a 30-person business, this looks familiar. A ChatGPT licence the operations manager tried for a week. An automation tool the founder saw at a conference that nobody touched after the first month. A CRM integration that was supposed to save hours but created more confusion than it solved.
The money isn't the issue. Most of these tools cost less than a team lunch. The real blocker is that nobody knows what to do with them, because there's no defined process for them to plug into. If quoting runs through email chains and spreadsheets, where does workflow automation fit? If your order-to-delivery process lives in someone's head, AI can't optimise what isn't documented.
And the businesses that did try? A good proportion got limited or disappointing results. Not because the tools were wrong, but because the tools had nothing structured to work with. We see this regularly: a business finds a piece of software through a video or an article, buys it, and then it sits on the shelf because nobody knows where to start changing or connecting it to existing workflows. Sometimes they use a few features for a while, then abandon the software entirely because they can't see it moving the needle on the issues they were trying to solve.
The Deloitte report calls this the gap between access and activation. At enterprise scale, activation means change management programmes and cross-functional rollouts. At SME scale, it means something simpler and harder: you need a structured quote-to-cash process before a tool can improve it. You need workflows that exist in a system rather than in someone's memory.
The tool is never the starting point. The process is.
84% of companies haven't redesigned work around AI. Most SMEs haven't designed it at all.
This is the statistic from the Deloitte report that should get the most attention and gets the least.
84% of companies surveyed have not redesigned jobs or the nature of work itself around AI capabilities. Despite high expectations for automation. Despite the fact that 36% of those same companies expect at least 10% of jobs to be fully automated within a year.
When asked how they are responding, 53% said they are educating employees to raise AI fluency. Only 30% are reimagining how their organisations work based on new AI-driven work patterns. Only 19% are redesigning career paths.
The majority response is training. Teach people how to use the tools. That's the beginning and end of the talent strategy for most companies.
For enterprise organisations, this is a missed opportunity. For SMEs, it's worse. Large companies at least have defined roles and documented processes that could be redesigned. Most £5m businesses don't. Work is shaped by who was hired, what they figured out on the job, and what the founder delegated last Tuesday. There's nothing documented to redesign.
What business owners want from AI is clear: automate the repetitive work, speed up the processes that drag, and free up time for the work that generates revenue. But you can't automate a process that doesn't exist in any documented form. You can't speed up a workflow nobody has mapped. And you can't free up time if nobody knows where the time is going in the first place.
Training people on AI tools without restructuring how work gets done is the equivalent of buying a faster car for a road with no lanes and no signs. The car isn't the constraint. The road is.
The real prerequisite is operational. Map your workflows and build accountability into the structure. Define who does what, when, and what happens when something falls between people. Then you can see where AI fits, because you can see the work clearly enough to know which parts a tool could handle.
Without that foundation, AI fluency is just another training initiative that doesn't change anything.
Only 21% have mature AI governance. But governance isn't what you think it is.
Deloitte found that only 21% of companies report having a mature governance model for autonomous AI agents. At the same time, 74% plan to deploy agentic AI within two years. The tools are arriving faster than the oversight.
When the report asked about AI risks, the answers were telling. Data privacy and security topped the list at 73%. Legal, intellectual property, and regulatory compliance followed at 50%. Governance capabilities and oversight came in at 46%.
These are enterprise concerns framed in enterprise language. Cross-functional governance structures. Audit trails. Compliance frameworks. For a founder running a growing business, that language doesn't connect.
But the underlying principle is the same at any scale. Do you know what AI tools your team is using? Do you know what data is going into them, especially client data? Is anyone accountable for the output, or are people using AI-generated results without checking them?
For a 30-person business, governance looks like this: a clear record of which AI tools are in use across the team, a policy on what client data can and can't go into them, and someone who owns the decision about when to trust AI output and when to verify it. A dashboard and a weekly review, not a 40-page policy document.
The risk of skipping this is already playing out. Teams start using AI ad hoc. They feed client data into free-tier tools with no data processing agreements. They make decisions based on AI outputs nobody is reviewing. In a growing business with no operational controls and visibility, this is already happening, and the founder usually doesn't know about it until something goes wrong.
Governance at SME scale is operational control applied to AI. If you already have a structured way of tracking what your team does and how decisions are made, extending that to AI tools is a small step. If you don't have that structure, AI governance is just one more thing on a list of things nobody is managing.
The gap between "we have AI" and "AI is working" is operational structure
Three gaps. Activation: tools bought but not embedded. Work redesign: 84% haven't restructured how work gets done. Governance: 21% have mature oversight. All three share a root cause.
Operational foundations must exist before any tool can deliver value. That is the real barrier to AI adoption for small businesses, and it has nothing to do with the technology.
Most founders who feel stuck with AI aren't lacking ambition or budget. They are lacking a starting point. And the starting point isn't technology. It's knowing how your business operates today, in enough detail that you can see where a tool would make a difference.
Think about it at the most basic level. If your customer data lives in three different spreadsheets, your invoicing is in Xero, and your sales pipeline is in someone's email or head, where does the AI tool point? The data must be in one place, cleaned up, and owned by someone before any AI application can work with it. Same principle that applies to enterprise data storage and practices, just at a smaller scale, with less margin for error.
Deloitte's own recommendation backs this up. Organisations that design for deployment from the outset, rather than treating scale as an afterthought, see far higher adoption. For a £5m business, designing for deployment means structured workflows, clear accountability, visibility over what's happening, and processes that exist in a system rather than someone's inbox.
What worked at £1m breaks at £5m. What worked at £5m breaks at £15m. AI doesn't fix that structural problem. It amplifies it. If your operations are held together by workarounds and tribal knowledge, AI will automate the workarounds and scale the confusion.
The businesses that will get value from AI in the next two years aren't the ones that adopt the most tools. They're the ones that fix the operations underneath first.
Fix the operations first. Then AI has something to work with.
The Deloitte report is written for companies with hundreds of employees and dedicated AI teams. The core finding applies to every growing business: when it comes to AI adoption, operational readiness is most likely the constraint, not the technology.
For founder-led businesses between £1m and £25m, the question isn't which AI tool to buy. It's whether the operations underneath are ready to make any tool useful. That means defined workflows, clear accountability, reliable data, and visibility over what's happening.
If your team is busy but nothing is getting more efficient, adding AI tools won't change that. If what worked at £1m is breaking at £5m, AI will accelerate the breaking.
The operations come first. Then AI has something to work with. In practice, that means mapping how the work actually moves through the business, assigning ownership at each stage, connecting the systems that need to share data, and then identifying where AI adds value. It is not a technology project. It is an operational one. We explain what that operational foundation looks like in practice in our guide to why every growing business needs an operating system.
If you're running a growing business and want AI to deliver real results, start with the operational structure.
We help founder-led businesses fix the processes, systems, and workflows that need to be in place before any technology can make a difference.
Stats Referenced in This Post
84% of organisations are increasing their AI investments. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
37% of companies are using AI at a surface level with little or no change to existing processes. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
3,235 business and IT leaders surveyed across 24 countries, August-September 2025. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
Workforce access to AI grew by 50% in one year, from under 40% to around 60% of workers with sanctioned AI tools. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
Fewer than 60% of workers with AI access use it in their daily workflow, unchanged from the previous year. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
84% of companies have not redesigned jobs or the nature of work around AI capabilities. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
53% of companies say their talent response to AI is educating employees to raise AI fluency. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
36% of companies expect at least 10% of jobs to be fully automated within a year. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
Only 30% of companies are reimagining organisations based on new work patterns from AI. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
Only 19% of companies are redesigning career paths in response to AI. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
21% of companies report having a mature governance model for autonomous AI agents. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
74% of companies plan to deploy agentic AI within two years. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
Data privacy and security is the top AI risk concern at 73%, followed by legal/regulatory compliance (50%) and governance capabilities (46%). Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
Only 20% of organisations are achieving revenue growth through AI today, compared to 74% who hope to. Source: Deloitte, State of AI in the Enterprise 2026, January 2026.
Sources