InsightsAI Insights
AI Insights

The agents are arriving inside software you already pay for. Here's how UK firms should switch them on

For most UK businesses, the first AI agent they use won't be something they went out and bought. It will turn up as a new tab in the finance system, the helpdesk or the ERP they already pay for. This week showed how quickly that is happening.

By StJohn Krog, Founder & MD

AI Insights27 October 2026

On 26 October alone, Oracle, NetSuite and Zendesk all announced agents that don't just answer questions. They monitor work, prepare actions and carry them out within rules you set. Meanwhile the UK's data regulator has made it clear that "the agent did it" won't count as an excuse.

That is good news. It means agents arrive next to your real data, inside systems your team already understands. But they still need switching on with care.

What launched this week

  • NetSuite SuiteAgents. At SuiteWorld, Oracle NetSuite announced industry-specific agents for software, product, services, healthcare, construction and nonprofit businesses. They are built to spot issues, recommend next steps and support multi-step execution "within established workflows". The product agent, for example, looks at inventory exposure and supply changes. The construction agent turns estimates into job budgets and prepares payment applications. The software, product and construction agents are available now in NetSuite Next, and the others are planned for a later release. (PR Newswire, 26 Oct 2026)
  • Oracle Fusion Agentic Applications. Oracle launched 25 agentic applications across finance, HR, supply chain, sales and service, including agents for reconciliation, procurement spend and payroll variance checks. The wording in the announcement is telling: almost every agent "prepares" actions "for review and approval". (Oracle, 26 Oct 2026)
  • Zendesk Employee Service AI Agents. Zendesk introduced agents that handle internal requests from start to finish, such as access approvals, laptop lifecycle management, employment letters and badge access. They run on administrator-built workflows called Advanced Action Flows. Zendesk says 300 early-access customers ran 6 million workflow executions in five weeks, saving an estimated 500,000 hours. That is the vendor's own estimate, so treat it as an upper bound. (Zendesk, 26 Oct 2026)

Closer to the high street, BT Business launched "AI from BT" on 1 October. It includes an AI receptionist aimed at the £3.7 billion a year BT says UK small businesses lose through missed calls. (BT, 1 Oct 2026)

Why this suits smaller firms

Most of the cost of a standalone AI project goes on plumbing: moving data out, keeping it in sync, and securing a separate tool. An agent built into your ERP or helpdesk skips most of that, because it already sits next to your records, permissions and audit trail.

The work these agents target is also where UK businesses are already seeing the effect. The ONS says 30% of UK businesses used at least one AI technology in September 2026, up from 9% in September 2023. Among AI users with ten or more staff, 38% said administrative and clerical roles were the most affected, up from 29% when the question was first asked in December 2025. (ONS BICS, 8 Oct 2026)

Reconciliations, stock checks, onboarding requests and missed-call follow-ups are exactly the kind of repetitive, rules-based work these agents are built for.

The confidence gap is real

Adoption is running ahead of know-how. Simply Business found 47% of UK small businesses now use AI tools, up from 22% in 2025, but only 19% feel "very confident" using AI day to day. (Simply Business, 8 Sep 2026) BT's own research found that 56% of small business owners say they are not yet getting the full value from AI. (BT, 1 Oct 2026)

An agent that can act, not just suggest, makes that gap matter more. A chatbot that gives a wrong answer wastes a minute. An agent that sends a wrong payment application to a client wastes a relationship.

The regulator is watching agents specifically

On 8 October the ICO opened a call for evidence on how data protection law applies to agentic AI. It closes on 20 November 2026 and covers security, transparency, accountability, automated decision-making, fairness and lawful processing. (ICO call for evidence) Announcing it, the ICO's Richard Nevinson said: "the fact AI agents act with autonomy is not an excuse for poor compliance." (ICO, Oct 2026)

The ICO's earlier research is just as direct: organisations "remain responsible for data protection compliance of the agentic AI they develop, deploy or integrate". (ICO Tech Futures: Agentic AI) If your supplier's agent touches personal data, you are still accountable for it.

A five-step switch-on plan

  1. List what's already in your stack. Check release notes for your ERP, CRM, helpdesk, accounting and ecommerce platforms. Many firms already have agent features available that nobody has looked at yet.
  2. Pick one boring, high-volume process. Choose something like month-end reconciliation, stock alerts or new-starter access requests. Success should be easy to measure in hours or errors, and a mistake should be cheap to undo.
  3. Start in "prepare, don't act" mode. Use the review-and-approve setting the vendors are clearly designing for. Let the agent draft and a person approve for the first few weeks, then widen its authority step by step.
  4. Write down what the agent can touch. Record which data it reads, which actions it can take, who approves them and where the log lives. One page is enough. It is also the evidence the ICO is asking organisations about.
  5. Check the costs before you scale. Agent features are often priced on credits or usage. Model what the cost looks like at ten times your pilot volume before you roll it out to every team.

Why this matters for Clarity Growth

At Clarity Growth we help UK businesses get value from the systems they already run, and built-in agents are the newest part of that. We help clients work out which agent features are worth switching on, connect them to the rest of the stack (website, ecommerce platform, CRM and finance), and set up approval steps, permission maps and logs so autonomy grows safely. If your software has started offering agents and you're not sure where to begin, that is a good conversation to have before peak season and before the ICO's guidance lands.

Sources

Related services

AI & automation

Automation

Technology consultancy