Choosing An AI Agency In London: What Gartner’s Warning On Vendor-Built Agents Means For Financial Firms

70% - Enterprises expected to abandon vendor FDE-built agentic AI by 2028 (Gartner, reported 5 October 2026)  ·  Under 20% - FDE engagements expected to turn recurring needs into core product capabilities (Gartner)  ·  9-10 months - Typical return on data extraction, payables and report creation (Gartner survey, reported 24 September 2026)  ·  26% - Agentic AI users unable to detect unauthorised agents (EY, 15 September 2026)

What did Gartner predict about vendor-built agentic AI?

Forward-deployed engineering is the model in which, in Gartner’s words, “vendor engineers work directly with customers to build and deploy solutions.” It has become a popular way for AI vendors to win business: put engineers inside the client, build something that works, then move on.

Gartner predicts that by 2028 “70% of enterprises will abandon agentic AI built by vendor forward-deployed engineering (FDE)”, “trapped by soaring costs and unable to evolve it on their own”, Digital Terminal reported on 5 October 2026. Gartner also expects that “less than 20% of FDE engagements will turn recurring customer needs into capabilities in the vendor’s core product”.

It added a warning about labels. Gartner described “a growing risk of ‘FDE washing,’ where consulting services are marketed as FDE.” Mukul Saha, a senior director analyst at Gartner, said: “FDE success starts with getting the engagement structure right, from scope and incentives to governance, ownership, and exit”.

What does an AI agency in London actually do for a financial firm?

The label covers very different businesses. Some AI agencies build marketing content or chatbots. Some resell a platform and configure it. Some, like us, build and run agents for one kind of client. For a UK financial firm, the useful question is narrower: will this agency take one manual workflow, build something that prepares it reliably, and leave the firm able to run it?

That last part is where Gartner’s warning bites. An agent built inside a firm by engineers who then leave can become a system nobody can explain or change. In a regulated firm, a system nobody can explain is also a governance problem, because a named person has to be accountable for what it does.

What should a financial firm agree before an AI build starts?

Gartner’s own recommendations make a good checklist. It advises firms to “Establish key contractual requirements beyond procurement’s scope, including deliverables, knowledge transfer, intellectual property rights and transition or exit responsibilities.” Translated for a 10 to 250 person financial firm, that means agreeing six things in writing:

1. Ownership. Who owns the code, the prompts, the configuration and the test cases at the end of the engagement.
2. Knowledge transfer. A runbook your own team can follow, and a handover session before the agency steps back.
3. A manual fallback. How the work gets done if the agent is switched off tomorrow.
4. Human oversight. Which outputs a person approves before they leave the firm. Gartner says each stage should “determine the appropriate balance between autonomy and human oversight”.
5. Measurement. A baseline of staff time agreed before the build, so the result can be checked against something real.
6. Exit. What happens to data, access and code when the engagement ends, and how long that takes.

Where does AI pay back fastest in finance teams?

Gartner’s survey of 160 senior finance function leaders, carried out from January to April 2026 and reported by CPA Practice Advisor on 24 September, found that data extraction, accounts payable and receivable automation and report creation “generally deliver returns within nine to 10 months”. Data management, insight generation and forecasting took longer to show value.

The pattern matches what we see in UK financial firms. The quickest wins are high-volume, rules-heavy preparation jobs: the daily mismatch report, transaction report checks and MI and board packs. They are easy to measure, easy to check and easy to switch back to manual if needed.

Why do governance and third-party risk matter more for AI agents?

An agent built by a third party is still the firm’s responsibility once it is running. EY’s survey of 202 senior AI leaders at large US companies, published on 15 September 2026, found that 26% of agentic AI users cannot detect unauthorised agents running internally and 47% have skipped their governance process for urgent deployments.

Financial standard setters are pointing the same way. The 2026 IIF-EY global survey of AI use in financial services, published on 17 September, reported that “Third-party AI models continue to challenge governance and validation efforts.” The Financial Stability Board published finalised sound practices for responsible AI adoption on 12 October 2026, and its Chair wrote that they “have been designed as a toolkit, adaptable to the size, type, and level of AI adoption within financial institutions.”

In practice, a workflow automation agency working with a regulated firm should expect to fit into the firm’s outsourcing and third-party risk process, not sit outside it. The agents themselves need the controls we describe in our guide to AI agents for financial firms.

What questions should you ask an AI agency in the UK?

• Which financial firms’ workflows have your team built before, and can we see the evidence pack from one?
• Do your agents ever change a system of record, send anything externally or move money without a person approving it?
• Where is our data processed, and is it ever used to train models?
• What will we own at the end, and could another team run it without you?
• How will you measure the result, and against what baseline?
• What happens if the agent is not live when you said it would be?

If a firm is also weighing a large-bank style rollout, our article on Barclays’ Claude rollout and what it means for smaller firms looks at the difference in scale.

How does BraivIQ approach ownership and exit?

BraivIQ is an AI agency in London for UK financial firms with 10 to 250 staff. Every engagement starts with a 14-day Proof Run on the firm’s own exports, read-only, which times one workflow on real cases and produces a baseline the firm signs. Every agent goes live with a named owner on the firm’s side, a runbook, a manual fallback and a written evidence file.

Our agents prepare the work and a person approves it. They never give advice, make investment or client decisions or move money. The senior team behind BraivIQ has built systems for one of the world’s largest trading floors and six UK banking brands.