Work
The systems running now, and the record behind them. Each study says what the work was, what it produced, and what I would do differently.
The estate
-
AI-assisted deal authoring inside a revenue-operations platform
A catalog-driven revenue-operations platform where a seller turns a plain-language brief into a structured deal. The model sits inside the workflow rather than beside it as a chatbot.
-
The Foundry: a machine that turns an idea into a live repository
The Foundry is line zero of the practice. It generates business ideas from market signals, triages them on a board, and provisions the winners as fully scaffolded repositories that coding agents can build and operate.
-
A private AI appliance that runs entirely on hardware you own
A self-hosted appliance running open models locally, powering always-on automations and, until October 2026, retrieval over a private corpus, without a call to a public API.
-
A real-time attention index fused from many public signals
A scheduled pipeline that fuses six independent public data sources into a single configurable interest score across nearly a hundred entities, refreshed continuously.
The record
-
Europe Delivers: reading twenty-nine countries with machines and analysts together
In 2019 my Singapore company shipped a thematic narratives study across twenty-nine European countries, built on a hybrid human and machine method, with all processing done in-browser and no data sent to a server.
-
Measuring hate speech in a language the machines could not read
English hate-speech lexicons existed. A robust Hindi one did not, and the code-mixed speech Indians actually write online was worse served still. The answer was a system that used the lexicons it had to find the terms nobody had written down yet. It was commissioned and it ran; the findings are confidential.
-
When a company can no longer be believed about its own good works
Corporations had lost the standing to describe their own contribution: the more they said, the less it counted. The answer was to stop making the claim alone, and to build the coalition that could make it credibly. The work was commissioned; what followed is confidential.
-
Making a cross-sector coalition something a company could actually join
India 2022 was a business-led coalition with an ambition that no company could sign. I worked on the architecture that turned it into four tracks, each with an owner, a government target and a pilot in the field.
-
The second wave is not the first wave again
Wave one asked leaders to invent the organisation's future. Every wave after it had a different job: getting people to catch up with a future already imagined. Most change programmes miss that, re-run the invention workshop forever, and wonder why it does not scale.
-
A stack is a training schedule that happens to include software
I was Chief Digital Officer, and the digital strategy I wrote and executed for the organisation refused the obvious move. Customer interaction was the differentiator, so it was the one thing not to automate. What got scheduled instead was the team's ability to run the tools.
-
The counting failure underneath India's malaria failure
A stakeholder audit that found malaria incidence in India under-reported by ten to a hundred times, and showed how the reporting gap starves the funding that would have closed it.
-
Answering a brief with the three questions the client had actually asked
The brand wanted to be more than a shampoo. Rather than arrive with a campaign, the deck named the three things we had heard the client say, turned each into a question, and proposed how we would answer them together.
-
Connecting influence to sales, for a brand that had never been able to
The pitch opened by admitting that the agency had never connected influence to results in the real world. Content to Commerce was the mechanism built to fix that, it was put to Pampers in China, and it ran.
-
Listening to Japanese mothers, with the sample size written down
Social listening, influencer analysis and review mining turned into a standing instrument for Pampers in Japan, built on a six-phase framework with durations, sample sizes and a confidence interval attached.
-
An agency model rebuilt around influence, six weeks before the contract ended
Six weeks from the end of the contract, I helped redesign how a global prestige skincare brand bought agency work across four Asian markets. The diagnosis came from search data, and the model was rebuilt so the fee paid for what the brand now needed.
-
A national communications architecture for a foundation's largest country program
I won and led the India communications mandate for the world's largest private foundation. The hardest part was the measurement framework, because the benchmarks it needed did not exist yet.
-
Seventy-four per cent of the complaints were not about the products
The brief was to shift the online conversation from negative to positive. The sampling said the conversation was a symptom: three-quarters of complaints concerned after-sales service, which is not a thing a campaign can repair.
-
A brand whose online presence was a complaints desk
Almost everything said about the brand online was news, a sales offer or a service problem. The work was to run the complaints as an operation with numbers attached, and to build brand conversation in the space that left.
-
Making an American brand Indian, without pretending it was not American
The brief was that Coca-Cola was not seen as an Indian brand. The answer was not to argue that it was, but to make its Americana something Indians could adopt, and the way in was to work out what happiness means in a collectivist country. I led the pitch and we won it.
-
Letting customers prove the product, when the brand had stopped being believed
Sony was losing an argument it kept trying to win by asserting. Across a laptop, a phone and a camera system, the same move worked: hand the proof to the people who already owned the thing, and let the evidence be theirs.
-
Giving graduating students one last thing they could do about it
The brief was to turn a graduating show into reach. The insight was about the students rather than the audience: people want one last chance to affect an outcome, and the moment before it closes is when they will work hardest on your behalf.
-
Sixty thousand organ donors in nine days
Organ donation in India was blocked by myth rather than by indifference, and India had 0.08 donors per million people. The campaign was built to convert belief into a signature, and the signature into a referral.