Field notes

Thinking from
client work

Short, practical observations from the rooms we work in. For the full conversation, follow David on LinkedIn.

May 2026
Agents

Start with the agent that checks

One approach that works well as we scale AI with clients: before building the agent that creates, build the agent that reviews. A document checker that reviews reports against a defined checklist. An agent that reconciles key numbers in support papers back to the source. A control layer that flags missing data points. It is easier to scale when you aren't constantly second-guessing the results: prove the review works, and the rest follows more smoothly.

Continue on LinkedIn
Apr 2026
Adoption

The Claude momentum, used well

Teams are moving much faster from idea to something people can actually use. That pace is exciting, and it rewards attention: where your data is going, what you are allowed to use, and how quickly costs build through subscriptions and token usage. The teams getting the most value work in close partnership with their tech teams, building in a way that integrates properly with the organisation's wider systems and ecosystem.

Continue on LinkedIn
Jan 2026
Copilot

Give Copilot another try

It has improved a lot: better conversations, handling of large documents, stronger analysis. The more interesting question now is where it is actually worth using. When teams are given time, context and focused training, they identify those opportunities themselves, and adoption and value creation follow naturally.

Continue on LinkedIn
Dec 2025
Reflections

Real value starts small

Fluency is uneven: executives often experiment first, while daily personal AI use elsewhere takes support to translate into value at work. The most consistent wins come from teams understanding the tools they already have and using them to solve immediate, practical problems. When teams are given space to learn, the impact goes beyond time saved: curiosity grows, and people find new ways to work on their own initiative. That shift in mindset is where momentum really builds.

Continue on LinkedIn
Nov 2025
Governance

The 30-second rule

Cyber attacks often succeed in the first 30 seconds because people act under pressure. The same applies to AI: a long document lands, you are under pressure, and a personal AI account can process it instantly. The tools are fast, and our emotions are faster. This is exactly where data governance and privacy practices matter most: as habits that hold up when things get busy, well beyond rules on a page.

Continue on LinkedIn
Nov 2025
Finance

Buried in budget season

Inboxes and Teams fill with budget files, comments and slides. What works for clients right now are focused AI automations that take some of the load away: data consolidation that pulls files, structures fields and flags gaps; a variance helper that turns "what changed" into plain English; a deck builder that creates or reviews budget packs. If you are in the thick of budgeting, it is a great moment to test where AI can take the pressure off.

Continue on LinkedIn
Oct 2025
Everyday AI

Everyday AI has so much to give

Finance and operations teams are using AI to compare data between systems and source documents, spotting inconsistencies and improving accuracy. The reduction in errors and manual checking is great, and the best part is watching people become more confident and curious, finding new uses in their day-to-day work on their own initiative. There is so much potential in the tools you already have: start there.

Continue on LinkedIn

Follow David on LinkedIn →