Skip to content

Speaking & podcasts

AI spend is becoming a number someone has to defend

As AI companies move toward public markets, their economics stop being a private matter — and so do their customers’. Token spend is turning from an engineering curiosity into an audited line item that boards, auditors and journalists ask about. Most organisations cannot answer. I build the tool that does, and I talk about why the gap exists.

Email hello@tokenok.io with the event or show, date, audience and format. Remote anywhere; in person around Sydney.

Talks

Nobody can tell you what their AI costs

Most organisations can produce a vendor invoice and nothing else — not cost per team, per project, or per outcome. Why the measurement gap exists, what it hides, and what it takes to close it without instrumenting every application.

Keynote · 20–40 min

The seller shouldn't count the savings

Advertisers stopped accepting publisher-reported impressions, and that produced IAB standards, MRC accreditation and third-party verification. AI spend is in the pre-verification era right now: the party recommending the optimisation is the party reporting the saving. What the adtech decade predicts about this one.

Keynote · panel · podcast

Your subscription is not free, and it isn't spend either

Flat-fee plans are the largest unmeasured pool in AI. Counting plan usage as dollars saved is wrong; ignoring it is worse. How to express plan usage as value delivered without inventing money, and why almost every tool in the category skips this entirely.

Panel · podcast · workshop

Not every job needs the frontier model

Routing work to the cheapest model that actually clears the bar, escalating when an answer comes back weak, and measuring the result. Live, across 37 models and 11 providers — including how a reasoning dial changes the economics, because thinking bills as output.

Technical talk · demo

The agent that spent $6,000 overnight

Runaway automated spend is a governance failure, not a bug. What the widely-reported incidents have in common, why budgets alone don't catch them, and what a kill-switch has to see to work.

Keynote · panel

Local-first as a procurement advantage

Building so that customer data never leaves their machine — what it costs you in features, what it buys you in sales cycles, and why 'we don't want your traffic' is the shortest answer to a security questionnaire.

Technical talk · podcast

Why this lands right now

Every public listing in this sector forces disclosure of unit economics, and every disclosure makes the same question louder for everyone buying AI: what are we actually spending, and what are we getting for it. The honest answer in most rooms is that nobody knows — the invoice is the only artefact. That is a good topic for an audience, because it is a problem they recognise in themselves rather than a product pitch.

What I bring

  • • A working product with real numbers — live rankings across 37 models, and a measurement method that labels every figure as provider-reported or estimated.
  • • A background in digital advertising, where third-party ad serving and verification solved exactly the conflict of interest now appearing in AI cost reporting.
  • • A position that is useful to an audience even if they never use the product: how to measure AI spend honestly, whoever they buy from.
  • • Demos that work live, including how a model’s reasoning setting changes what a job costs.

Programming an event or booking a show?

Tell me the audience and I’ll tell you honestly whether I’m the right person for it.

hello@tokenok.io