1 Year. 33 Workshops. 366 Agents: What We Found.
We spent twelve months asking financial services teams a different question. Not what AI could do. What was actually wasting their time.
Zandra and her team at Zygens
Twelve months ago, I started asking financial services teams a simple question.
Not "what could AI do for you?" Not "where do you see AI fitting into your strategy?" Those questions produce answers shaped by what people think they're supposed to say.
I asked something more direct: what takes the most time for the most people, and why?
The answers were not what we expected. Not because they were surprising in isolation. But because they were identical, across organisations that had nothing else in common.
The numbers
In the twelve months to March 2026, Zygens ran 33 AI agent design workshops with 330 individuals across financial services, legal, logistics, health and e-commerce. Together, those participants designed 366 AI agents, each grounded in a real operational pain point, scored for value and risk, and mapped to a deployment priority.
Financial services dominated. Sixty-four percent of workshops. Fifty-eight percent of all individual training hours. Two hundred and seventy of the 366 agents designed.
That concentration was not planned. It reflected demand.
The organisations pushing hardest to move from AI ambition to something deployable were, consistently, in regulated financial services. Insurance. Alternative lending. FinTech. Credit unions. Building societies. Not because they were the most enthusiastic about AI. Because they were under the most pressure to get it right.
What the urgency is actually about
The FCA and Bank of England's joint survey found that 75% of UK financial services firms were already using AI in some form by late 2024. The figure sounds reassuring. It isn't.
Adoption breadth is not deployment depth. Most of those organisations are using AI in isolated pockets, not in the core operational workflows where the real costs live.
McKinsey puts it plainly. AI pioneers in banking are building a 4% ROTE advantage over laggards. That gap is compounding now.
The organisations in our programme understood this. They weren't attending workshops to learn what AI was. They were attending because they had committees formed, strategies written, pilots completed, and nothing in production.
That picture is consistent with the wider industry data. Sixty-five percent of financial institutions experience implementation delays averaging 14 months. Only 29% report that AI has delivered meaningful cost savings to date.
The problem is not capability. The technology works. The problem is sequencing, governance immaturity, and the absence of a structured way to move from "we should be doing this" to "this is running."
What we actually found
When you ask 330 people across financial services to describe their biggest operational frustrations, ten themes emerge. They appear regardless of sub-sector, organisation size, or seniority.
The same problems surface in insurance underwriting teams and in collections businesses. In FinTech product teams and in building society compliance functions.
Information that exists but is practically inaccessible. Reporting cycles that consume analyst days. Compliance monitoring that is manual, repetitive, and error-prone. Correspondence that follows predictable structures but still requires hours of human drafting time. Document data extraction that should have been automated years ago.
None of this is glamorous. None of it appears in AI strategy presentations. All of it is where the time actually goes.
More than half of the 366 agents designed across our programme landed in what we call the OPTIMISE quadrant of the Zygens DDAD Framework: a scoring model that maps each agent against operational value and deployment risk. OPTIMISE means high-value, low-risk, internal-facing, deployable now. A further 22% landed in INNOVATE. Together, nearly three quarters of all agent designs were assessed as low-risk.
This matters. There is a persistent belief in leadership teams that starting with internal efficiency agents is deferring the real opportunity. The data says the opposite. Organisations that build confidence and governance capability through these deployments are significantly better positioned when they move into customer-facing and decision-making territory. Those that skip this phase face higher failure rates, greater regulatory exposure, and the reputational risk of getting customer outcomes wrong at scale.
The 52% of agents in OPTIMISE are not the consolation prize. They are the foundation.
What this series covers
Over the next several months, we will publish the full findings from our programme: what teams are trying to fix, how they are prioritising, where governance design is making or breaking deployments, and the five agent categories that consistently emerge as the strongest candidates for rapid deployment.
Every piece draws on the same dataset: 33 workshops, 330 individuals, 366 agents. Not theoretical. Not modelled. Derived from what real teams in regulated organisations are actually trying to build.
The full report is available to download now. The live programme data is on a live dashboard.
What to do next
If you are in financial services and recognise any of this, the Discovery Programme is where to start.
It is a structured, facilitated engagement that maps your highest-value AI opportunities, scores them for risk and complexity, and produces a prioritised deployment roadmap. It takes weeks, not months. It produces working direction, not a slide deck.
Book a Discovery Programme conversation at zygens.com or contact us at hello@zygens.com.
Zandra Moore MBE is Co-Founder and CEO of Zygens.
