News
5 min readAgentic AI adoption 2026: what the numbers say for small firms
Only 17% of organisations have deployed AI agents so far, according to Gartner. Here is what that gap means for a small firm deciding whether to act.

Agentic AI adoption 2026 looks very different depending on which report you read. If you run a firm of a few dozen people, you're being asked whether to act, and the headline percentages won't tell you what to do on Monday.
What agentic AI means in 2026, and why the numbers vary so much
An agent is software that takes a goal, works out the steps, uses tools such as your email or CRM, and acts without a person approving each move. A chatbot answers a question. An agent does something about it. That difference is the first thing to check in any report.
The numbers vary because surveys don't agree on terms. Omnius reports adoption figures that run from 10% of organisations with agents in production to 80% of responding UK financial firms with decision-support agents. Those figures measure different things. Before you quote one, ask what counted as an agent and what counted as adoption.
Agentic AI adoption 2026: how many organisations have deployed agents
Gartner says only 17% of organisations have deployed AI agents to date, yet more than 60% expect to do so within the next two years. Forrester says three-quarters of enterprise leaders are adopting agentic AI, but only a small minority have it running in meaningful production beyond what it calls 'agentish' chatbots.

Deployed, piloting and scaled are different things
For a buyer, the useful split has three stages. Deployed means something is switched on. Piloting means a test with a few users. Scaled means the agent runs real work across the business and people depend on it. Most headline figures blur these together.
Scaling is where the gap shows. A McKinsey figure reported by trydock.ai says 40% of companies with more than $1 billion in revenue are scaling AI agents, up from 27% a year earlier. Smaller companies are flat at 22%. That is a secondary source, so treat it as a direction rather than a precise measure. For a small firm, the point is simple: the large players are pulling ahead, and the smaller end has barely moved.
Why scaling stalls
Forrester says scaling fails on task complexity, not on how many agents you own. It also warns that stitching isolated agents together without a shared registry or routing leads to duplication and drift.
Cost is the other trap. Omnius reports that 46% of G2 respondents say agents cost more to run than planned. Buying licences isn't adoption either. SuiteBriar notes that if people aren't trained and shown how agents fit their work, usage drops off quickly.
How isolated agents turn into duplication and drift
Build
One agent per bounded jobClear goal and tool accessConnect
Shared registry of agentsRouting between agentsRun
Cost tracked against planOutput checked for driftWhere agents are working today
Gartner places agentic AI at the Peak of Inflated Expectations and says most deployments remain narrowly scoped. It adds that fully autonomous agents are not ready for the majority of enterprise use cases. In practice, the working deployments are small, bounded jobs: answering a standard enquiry, chasing an unpaid invoice, booking a slot, pulling a weekly report.
Our view is that the same rule applies to you. Start where the job has a clear start, a clear finish and a person who can check the result.
What small firms should copy from enterprise leaders
Copy the discipline, not the budget. Three things carry over:
- Pick one process with a measurable outcome before you pick a tool.
- Keep a written list of what each agent is allowed to do, and who approves anything beyond that.
- Train the people who work alongside the agent, and show them where it fits their day.
If you want a partner to scope a bounded flow like this, look at our AI systems and automation service, which covers assistants that handle replies, follow-ups and reports.
What small firms should avoid
Agent washing. SuiteBriar cites a Gartner estimate that only around 130 of the thousands of vendors selling agentic AI offer real agent capabilities. Many products are rebranded chatbots or automation tools. To check a product, ask three questions: what does it do without someone pressing a button, what can it change in your systems, and what happens when it gets something wrong. If the vendor can't answer in plain terms, you're probably buying a chatbot at agent prices.
Licences without a plan. A seat with no owner, no training and no defined task is a monthly cost with no return. Our guide on whether agentic AI is worth it for a small business sets out the questions to answer before you buy.
Open-ended spend. Set a ceiling before the first run. The example below shows how.
Is agentic AI outdated, or replacing itself?
The sources don't say agentic AI is outdated. Gartner's Hype Cycle places it at the Peak of Inflated Expectations, which describes where expectations sit, not a verdict that the idea has failed. Our view is that the label will keep changing. The question for an owner stays the same: does this agent do a named job, reliably, that someone would otherwise pay for?
Example: setting a budget ceiling for an agent
These numbers are made up to show the method. Inputs:
- Licence fee: £180 a month
- Usage charges: £120 a month
- Staff time checking results: 3 hours a month at £25 an hour, so £75
Planned monthly cost: £180 + £120 + £75 = £375.
Ceiling: plan plus a 20% buffer. £375 × 1.20 = £450.
Rule: if the monthly bill goes past £450, the agent pauses and someone reviews what it did before it runs again. The 20% is our rule of thumb for this example, not a figure from any survey. Before you commit, check the liability side too, because who pays when an agent gets it wrong matters as much as what it costs.
How to read the Hype Cycle and analyst surveys without being misled
Ask five questions of any report:
- What counted as an agent? A chatbot with a plug-in may be counted.
- Is it talking about deployed, piloting or scaled?
- Who answered? Enterprise firms say different things from firms of thirty people.
- Who paid for it? Vendor-sponsored surveys tend to favour the vendor's product.
- When was the data collected? Agentic AI is moving quickly, so older figures age fast.
If the report can't answer most of these, use it for direction only.
If you want to know where agents fit your firm, not a large enterprise, start with our free AI plan. It's a short look at your top tasks and what is worth automating first. If you'd rather talk it through, book a call and we'll go through it with you.
Sources
Questions people ask
How many companies are using AI agents?
Gartner says 17% of organisations have deployed AI agents so far, and more than 60% expect to within two years. Forrester reports that three-quarters of enterprise leaders are adopting agentic AI, but few have it in meaningful production. Figures vary because surveys define 'agent' and 'adoption' differently, so check the definition first.
What is the gap between deploying and scaling agentic AI?
Deploying means something is switched on. Scaling means the agent runs real work across the business and people rely on it. A McKinsey figure reported by trydock.ai puts scaling at 40% of companies with over $1 billion in revenue, but flat at 22% for smaller companies. Forrester says scaling stalls on task complexity and poor coordination between agents.
Are AI agents mature enough for small businesses?
Gartner says most deployments remain narrowly scoped, and fully autonomous agents are not ready for most enterprise use cases. For a small firm, that points to bounded jobs with a clear finish and a person checking the results, rather than broad autonomy. Whether it pays depends on your process and your running costs, which you can test on a small scale first.
Is agentic AI outdated?
The sources we used don't say so. Gartner places agentic AI at the Peak of Inflated Expectations, which is a point on its curve rather than a verdict that the idea has failed. The useful test for an owner is whether a named job gets done reliably and someone would otherwise pay for it.
Keep reading

Small business AI adoption: why owners are moving slowly
Surveys disagree on how many small firms use AI. Here is what the numbers mean, and how to pick one task, one metric and one review date.
Noah Hany · 16 Nov 2026

AI tools for solo business owners: what you actually need
A short, affordable AI stack for a one-person business, what to leave out, and when bringing in an agency is worth the money.
Noah Hany · 10 Nov 2026

Is agentic AI worth it for a small business right now?
Agentic AI pays off for some repeated, measurable tasks and not for many demo-friendly ones. Here is a simple test and the questions to ask before you sign.
Noah Hany · 10 Nov 2026