Technologist's playbook part 2: Durable AI projects

Last week I shared a few tips from a technologist’s playbook to help creative financial advice firms kick off their AI experiments on the right footing. That playbook runs deeper.

Here are three principles to make your AI projects durable and ready to scale:

👉 Treat QA as a first-class AI citizen

Advice firms often think about AI in terms of automation — structuring meeting notes, drafting advice documents, running compliance checks. But AI is at least as powerful for testing as it is for building.

You can use it to stress-test individual workflow steps, run bulk regression tests across dozens of scenarios, and validate the end-to-end experience from the adviser, client, or compliance manager’s perspective.

AI-powered QA means you can launch faster, with fewer surprises, and keep iterating with confidence.

👉 Build in observability from the start

“Observability” is a nerdy word, but it simply means: can you understand what’s happening inside your system by looking at its outputs?

For AI projects, that means:

• Logs: detailed enough to show merged prompts and responses, but scrubbed of PII
• Metrics: performance stats, accuracy rates, error counts, iteration counts for agentic flows
• Costs: token usage and spend tracked at a transactional level

Without these signals, you’re flying blind. With them, you can spot anomalies early, prove your ROI, and hold both your team and your vendors accountable.

👉 Don’t skimp on error handling and audit trails

Low-code prototypes and quick outsourcing projects often skip the boring parts: graceful error handling, clear failure messages, and reliable audit trails of who did what, when.

In wealth management, these aren’t nice-to-haves; they’re essential for explaining a system’s behaviour to a regulator, or to a client whose data you’re responsible for.

Bake them into your design early — or choose tools that provide them by default — and you avoid brittle workflows that leave you exposed when questions arise.

These principles aren’t glamorous, but they’re what separate a clever demo from a system you can trust with your advisers, clients, and regulators. 🙂