The tools worth watching.

Most AI announcements are noise. A few developments genuinely matter for financial services professionals.

Technology Perspective · Feb 2026

scroll

Signal vs Noise

Cutting through the noise

The volume of AI-related announcements is relentless. New models, new products, new capabilities — every week brings another wave. For anyone working in financial services, the challenge is not awareness. It is discernment.

What actually matters? What will change how we work? And what is marketing wrapped in a press release?

After spending considerable time separating signal from noise, here is where I am focusing my attention.

“The valuable AI applications in financial services aren't creative. They're precise, high-stakes, and auditable.”

Development 01

Structured reasoning in large language models

The most significant development in the past twelve months has not been any single product launch. It has been the steady improvement in LLMs' ability to follow complex instructions, maintain context over long conversations, and produce structured, reliable outputs.

Why It Matters

The valuable applications are not creative writing. They are precise, high-stakes tasks: summarising regulatory documents, extracting key terms, generating compliant communications.

The Threshold

The better models get at structured reasoning, the closer they get to being genuinely useful in a professional context where accuracy is non-negotiable.

This matters for financial services because the bar for reliability is exceptionally high. A model that hallucinates occasionally is a novelty. A model that follows complex regulatory logic consistently is a transformation.

“A tool that correctly extracts information from a mortgage application 95% of the time is a curiosity. At 99.5%, it's a transformation.”

Development 02

Real-time voice and conversation AI

The ability for AI to participate in real-time conversations — not just transcribe them, but actively support them — is maturing faster than most people in banking realise. The implications for branch environments are significant.

→Real-time prompts during conversations — a reminder to discuss a specific product feature, a flag that the customer's circumstances suggest an alternative product.
→Regulatory disclosure surfacing — a note that a particular disclosure needs to be made, triggered by the conversation context rather than a colleague's memory.
→Automatic interaction documentation — the conversation is captured, structured, and made audit-ready without the colleague writing a single note.

This is not science fiction. The underlying technology exists today. The question is deployment readiness in regulated environments — privacy, data governance, and colleague acceptance.

Development 03

Document understanding and extraction

Financial services runs on documents — applications, contracts, regulatory filings, compliance reports. The ability of AI to read, understand, and extract structured information from these documents is improving dramatically.

What interests me is not the headline capability but the reliability. We are approaching the threshold where automated document processing moves from experimental to production-grade.

Applications

Mortgage applications, credit assessments, KYC documentation — structured extraction that previously required manual review.

Regulatory

Policy documents, compliance filings, regulatory updates — summarisation and impact analysis at a speed no human team can match.

“The tools worth watching are the ones that make existing professionals more effective — not the ones that promise to make them redundant.”

The Noise

What I am ignoring

AI-generated images. Chatbots that try to be your friend. Autonomous agents that promise to replace entire job functions. These may have applications somewhere, but they are not where the near-term value lies for financial services professionals who need tools that are reliable, auditable, and compliant.

The filter is simple: does this technology make an experienced professional measurably better at a high-stakes task? If the answer is yes, it's worth paying attention to. If the answer is “it's impressive but I'm not sure what I'd use it for,” it's noise.

The tools worth watching are the ones that make existing professionals more effective — not the ones that promise to make them redundant.

The question isn't which AI tool to buy.

It's which capabilities will make your people genuinely better at the work that matters.