How Downing moved from AI intent to execution on Claude
Downing, a multi-asset-class responsible investment manager, wanted to grow without growing headcount one-for-one. Inference Group used Claude to find exactly where that was possible, then trained Downing's own AI Champions to build on Claude. Every team we worked with now has at least one AI solution using Claude in production, built by Downing's own people.
teams now have at least one Claude solution in production, built by their own champions
people trained to build on Claude at the firm-wide session
AI opportunities identified by Claude-powered analysis, 136 prioritised into a roadmap
What Downing wanted
Downing is a responsible, multi-asset-class investment manager, investing primarily across renewable energy, infrastructure, healthcare and education. The question it brought to us was about growth: if the business scaled quickly, how could it do so without a one-to-one increase in headcount?
Put simply: where, specifically, could AI let teams grow their output without growing the team? Downing did not want a report about it. It wanted the capability to build the answers itself, and to keep building after we left.
Finding where AI could work, with Claude
We ran Control Tower interviews across 12 teams with 35 participants. Using Agent Teams built on Claude Code, those interviews and workshops were assessed against a large repository of what works, surfacing over 150 opportunities. Very few were ruled out. They were deduplicated and turned into a prioritised roadmap, organised by who could deliver each one, so value could start flowing immediately rather than waiting on a single big programme.
That analysis was the means, not the outcome. The outcome was a queue of specific, buildable things, and a plan that put most of them in the hands of Downing's own people.
Training the champions to build on Claude
We trained Downing's AI Champions in the full use and functionality of Claude: Claude and Claude CoWork for everyday work, Claude Code for building, Claude in Chrome and the Office apps for the tools they already live in, and Claude Design for rapid prototyping. Two AI Champion sessions covered the art of the possible. A firm-wide session took 160+ people through it.
The training application itself was built in Claude Design, and used live as a demonstration of how quickly a working solution can be stood up on Claude.
The champions then trained their own teams. That is the point at which this stopped being our engagement and became Downing's capability.
What the champions built, and what runs today
Every one of the 12 teams we worked with now has at least one AI solution in production use, built on Claude by Downing’s own champions. Not prototypes. Not a pilot cohort. Working assets that colleagues use to do their jobs, every week.
Investment teams use Claude to work through incoming opportunities at a pace and depth manual review could not sustain, so more of each analyst's time goes on judgement rather than reading.
Relationship and record-keeping work that previously consumed hours of manual entry now runs on Claude, keeping the data current without the drag.
Champions build working tools as artifacts in Claude and hand them straight to their teams, turning a repetitive process into something a colleague can run in a couple of clicks.
Champions have built Claude skills that run on a schedule, so recurring work happens without anyone having to start it.
Alongside what the champions build, Claude runs across Downing as everyday working infrastructure: firm-wide adoption on Claude, Claude integrated into Downing’s own systems, and the prioritised roadmap delivered as a live web application with a Sonnet 5 conversational interface, so teams can question the roadmap, update it, and feed new opportunities in by document or voice recording for a Claude Agent Teams process to analyse. Inference Group continues to support the deployment through a Senior Advisor retainer and scaled Claude Code and CoWork user support.
A self-driving company: one where the people closest to the work build their own AI, on Claude, and keep building. Downing did not buy a roadmap and wait. Its champions were trained, they built, and every team we worked with is running what they built.
A roadmap organised by who can deliver
The point of the roadmap is not just what to build, but who builds it. That is why the champion-led column is the biggest one, and why it is the column that has already delivered.
Opportunities Downing's own trained AI Champions can build now, with growing knowledge and supporting tools. This is where the production solutions have come from.
Opportunities that need engineering support or touch the wider architecture and systems.
High-impact projects that benefit many across Downing, needing specific skill sets and sometimes a buy-or-build decision.
Baseline enablers other work depends on, such as data integration and governance like an AI policy, sequenced by Downing.
“Working with Inference Group gave us the clarity and momentum we needed. Their expertise helped shape our AI roadmap and set us up for long-term success. We're already seeing the benefits as our teams turn strategy into real delivery.”
Yes. As a result of the engagement, Downing's own people build and deploy Claude applications and workflows, and every team we worked with now has at least one live in production — from the roadmap application that runs on Claude to the day-to-day workflows their AI Champions have automated with Claude Code and the wider Claude tools.
Find where AI grows your business without growing the team
A 30-minute scoping call with the Inference Group team.