My Advice to Banks on AI: Omid Pakseresht of GOODFOLIO

I spoke with Omid Pakseresht, Founder and CEO of GOODFOLIO, an applied AI platform building industry-grounded AI systems for global midsize enterprises. With flagship products serving CPG, financial marketing compliance, and strategic intelligence, Omid brings a systems-thinking perspective to how banks should approach AI deployment and infrastructure.

Over to you Omid - my questions are in bold:

Can you give us an introduction to you and an overview of your organisation?

I'm Omid Pakseresht, founder and CEO of Goodfolio. With my co-founder Nima, we built Goodfolio as an applied AI platform - designing, building, and deploying good AI for global midsize enterprises, including financial services. Our flagship products, NuStream for CPG, finspector for financial marketing compliance and GoodMora for strategic and operational intelligence, sit on top of Sepanta, our internal six-layer agentic infrastructure that turns deep industry knowledge into deployable AI systems. We've been bootstrapped since 2023, generated more than £1m in revenue, and recently closed an oversubscribed pre-seed round.

If you were advising a bank CEO today, what would you say it the single biggest mistake they're making with data and AI?

Treating AI as a procurement decision instead of a systems decision. Most banks are buying point tools - a copilot here, a fraud model there - without designing how those pieces connect to data, decision rights, and the people on the front line. Without a system, things don't compound, and you end up with AI fatigue and a portfolio of pilots that never reach production. The mistake isn't the tool you picked; it's the absence of an architecture for how AI gets used, governed, and improved.

What's one AI or data capability banks should prioritise in the next 12–18 months, and why?

Industry-grounded agentic workflows over the long tail of internal work - the judgment-heavy, repetitive tasks that are too nuanced for RPA but too costly to keep on senior people. Marketing-compliance review, KYC remediation, model risk documentation, and credit-memo drafting are ideal first ground because the inputs, outputs, and cycle-time savings are measurable. The window matters now because retrieval, reasoning, and tool use can finally be combined safely - but only if the data and policy layers underneath are real. The capability to prioritise isn't 'an AI'; it's a deployable agentic stack with the guardrails and observability to put it on the critical path.

Where do you see banks overestimating AI, and where are they underestimating it?

Banks overestimate frontier model magic and underestimate the supporting architecture - the catalogues, policy layers, and orchestration that decide whether a model actually works in production. Everyone is impressed by demos and disappointed by deployments because the demo is 5% of the system. On the other side, banks consistently underestimate how much value sits in the hidden architecture of their own business: the unwritten rules, undocumented exceptions, and tribal expertise that AI can finally make visible and reusable. That's the upside most are missing - using AI to industrialise the institutional knowledge they already have.

What does "good" actually look like when AI and data are working well inside a bank?

Good looks like AI that's invisible to customers but deeply embedded in how work gets done - faster decisions, fewer handoffs, and a clean audit trail behind every output. It means eventually the same agentic infrastructure powers compliance, lending, and operations with unique customisable AI systems, instead of every team building their own toolkit and their own risk surface. Good AI is industry-grounded, accountable, and aligned with the regulatory and operating constraints of the bank, not a generic productivity layer bolted on top. When it's working, you don't notice the AI; you notice that the bank moves faster and breaks fewer things.

What's the hardest AI or data decision bank executives are avoiding right now, and why?

Whether to consolidate AI as a shared platform capability or let each business line keep buying its own tools. The honest answer is that distributed buying is creating ten flavours of risk, ten cost structures, and ten future migrations - but consolidating means hard conversations about ownership, budget, and which legacy systems get rationalised. Executives avoid it because it's a multi-year structural decision dressed up as a tooling question. The banks that pull ahead will be the ones treating AI as core infrastructure and capability now, designing for the next decade rather than the next budget cycle.

Thank you Omid! You can connect with Omid on his LinkedIn Profile and find out more about the company at www.goodfolio.com.