Writing · Automation and decision quality

How small can a T&S team get now that AI tools are this good?

I've been in a few conversations lately with leaders asking how small a T&S team can get now that the AI tools are this good.

The tools are good (great, even). Classifiers that used to flag any clip with a weapon in it can now tell game footage from real violence - and contextually tell you what’s happening in a clip. Models catch coordinated spam and scam patterns faster than a human queue ever could. A lot of review volume that needed people two years ago doesn't anymore.

The regulatory side has moved just as fast, in the other direction. The DSA requires qualified staff involved in appeals. Ofcom's codes under the UK Online Safety Act expect a named senior person accountable for illegal content, reporting and complaints, and they list properly funding and staffing moderation teams as a safety measure. Age assurance rules now differ in nearly every major market.

Each of those needs someone who understands the policy, the product and the regulator, and who can put their name on a decision.

Having built and run a T&S function, my view is to automate the clear-cut volume and put the savings into senior people who own policy, appeals, risk and emerging threats. That setup works for the business and holds up with regulators.

Cutting the team down to a singular person and a dashboard (in-house or outsourced) saves money in the short term. Then a regulator asks why the system made a specific call, and nobody inside the company can answer. Fines under the DSA can reach 6% of global turnover, and under the Online Safety Act up to 10% of global revenue. That's a lot more than the headcount you saved.

Related

  1. Since 2024, the job is proving it works

    Since 2024 the question has moved from whether you have a policy to whether you can show it works: UK risk assessments that must be redone before significant changes, age assurance as the foundation, 48-hour removal of intimate images under the US TAKE IT DOWN Act, and free open-source tooling. Those duties land on different teams, so the leadership team should name one owner for the whole picture.

    2 min read · Essay · What's changing

  2. Automation rate isn't a measure of maturity

    Automate as much as the evidence supports. Where a wrong decision can't be reversed or someone's safety is at risk, automation prepares the case and a person closes it.

    1 min read · Essay · Automation and decision quality

  3. Mexico's age-check decree: build to the strictest bar you face

    Mexico's presidential decree, in force since October 14, 2026, requires platforms, video games and app stores included, to use age checks that go beyond self-declaration, start new accounts on the strictest settings and give parents supervision tools, with a report due within 30 days. With no two markets asking for quite the same thing, build one approach to age assurance and child defaults to the strictest bar you face.

    1 min read · Essay · What's changing