AI Training · One day · In-house or open cohort

Built for the people who buy the science.

One intensive day on frontier AI agents for business development, licensing, search and evaluation, competitive intelligence and commercial teams. No code. You leave able to run a scouting question end to end, check what comes back, and buy the right tool with your eyes open.

Book a seat · $4,999 See the day

$4,999 per person. Corporate and group discounts for teams.

Who it is for

Scientists by training who want to understand the machine, not just operate it.

Heads and directors of search and evaluation, business development and licensing, competitive intelligence, portfolio strategy and commercial insights — and the people on their teams who will actually run the tools. The programme is vendor-neutral: everything taught works whether your company has standardised on Claude, on GPT, or on both, because the building blocks are now the same.

Run as an open cohort alongside the dinner series, or in-house for one company with your own assets and your own data as the material.

The day

The morning is the machine. The afternoon is your job.

Twelve modules and a live capstone, 09:00 to 17:45. Every module is built on work that has shipped inside a pharma company, and every claim is traced to a source you can check. No vendor demos.

  1. 08:30

    Doors, coffee, laptops open

  2. 09:0030 min
    Module 1 · The machine

    How a frontier model answers, and why it invents sources

    Context windows, tokens, reasoning effort, and the specific ways a confident answer goes wrong. When to trust a number and when to go to the source.

  3. 09:3030 min
    Module 2

    Tools: the agent's hands

    Function calling, web search and fetch, code execution, reading PDFs and files, citations. The same idea in Claude and in GPT, and what changes once the model can act rather than just write.

  4. 10:0030 min
    Module 3

    MCP servers: plugging your own data in

    What the Model Context Protocol is, why both Claude and GPT connect to it, and what it means to point an agent at your pipeline database, your document store or a data vendor. The questions your security team will ask, answered in advance.

  5. 10:30

    Break · 15 min

  6. 10:4545 min · hands-on
    Module 4

    Skills, projects and instructions: reusable playbooks

    Turning a scouting method into a folder of instructions the agent loads on demand — Skills in Claude, custom GPTs and projects on the other side. You write one for your own team in the room.

  7. 11:3030 min
    Module 5

    Deep research, agent modes, memory and context

    What changes when the agent runs for an hour instead of a minute: compaction, memory, long-running tasks, and the trap of more sources with less reliability.

  8. 12:0030 min
    Module 6

    Evals and verification: knowing whether it works

    Building a small evaluation set from decisions your team has already made, measuring completeness rather than eloquence, and the limits of letting a model grade a model.

  9. 12:3030 min
    Module 7

    Cost, controls and where it runs

    Token economics, effort levels, caching and batch processing; open-weight models on your own infrastructure against frontier APIs; data residency and what your procurement review will require.

  10. 13:00

    Lunch · 45 min

  11. 13:4545 min
    Module 8 · Your job

    Search and evaluation: asset scouting end to end

    From a therapy-area question to a candidate universe to a shortlist. Completeness as the only metric that matters, Chinese-language sources, and a test a vendor could fail.

  12. 14:3030 min
    Module 9

    Competitive intelligence with agents

    Trial landscapes, competitor tracking and monitoring pipelines. Moving the team from collection to interpretation, and what a machine-first CI report has to show before anyone acts on it.

  13. 15:0030 min
    Module 10

    Due diligence and the evidence file

    Sourced claims, the diligence memo a committee will accept, and where the accountability line sits when the analysis was machine-made.

  14. 15:30

    Break · 15 min

  15. 15:4530 min
    Module 11

    Portfolio strategy, launch and commercial intelligence

    Benchmarks and indication priorities; launch landscapes by market and brand tracking; and the EU transparency obligations that now apply to an AI-assisted competitor deck.

  16. 16:1530 min
    Module 12

    Reports, platforms and buying well

    When to commission a delivered report and when to run a platform yourself; metered credits and what they really cost; writing a vendor brief; and comparing vendors when every benchmark was built by a vendor.

  17. 16:4560 min · hands-on
    Capstone

    A live scouting question

    Your own question, run end to end with the tools from the morning, finished as a committee-ready shortlist and memo. You keep the playbook.

  18. 17:45

    Close and drinks

Pricing

One price per seat. Better rates for teams.

Attending from a pharma company? The seat value is stated on the invoice so you can clear it with your compliance team.

Why this exists

Nobody sells AI training to the commercial seat.

There is a great deal of AI training in pharma, and almost all of it is for R&D, IT, regulatory or manufacturing. The courses aimed at commercial teams are about field force and omnichannel. We looked for a programme built for business development and licensing and could not find one.

Meanwhile adoption among competitive intelligence teams rose sharply in the last year, and a large share of practitioners now use these tools daily. That is a population that was handed the tools without being taught what they do.

Book a seat, or bring your team.

Tell us who is coming and we will come back with dates, and a group rate if there are five or more of you.

Book a seat · $4,999 Ask about in-house