AI Sales Companion

6 Interactive Sessions

Short, interactive sessions: watch it work, try it, then build it yourself. You can go deeper at any time in the written chapters.

1. 1

   Gather every source in one place

   Everything a rep needs before a call is already written down, but it is spread across five systems in five formats. The companion's first job is to bring it all into one store, in one shape, where every fact still says where it came from.

2. 2

   Search by meaning, not by words

   You search for “pricing pushback”, and the call says “She'll want ROI projections”. No words in common, the same worry. A word search misses it. To find it, the companion cuts each record into chunks and turns each chunk into numbers that capture its meaning, so a question lands next to the chunks that mean the same thing.

3. 3

   Pick the few chunks that matter

   “What should I know before my call with Acme Corp?” could pull back dozens of loosely related chunks. Retrieval picks the few that matter: the deal, the CFO's request, the SAP setup, the people. It runs a meaning search and a word search, merges them with reciprocal rank fusion, and reranks by source, date and repeats.

4. 4

   Only say what a source supports

   The companion writes “Acme's budget is $200K.” Did a record say that, or did the model make it up? A rep who repeats a made-up number to a customer cannot take it back. The AI gateway is the set of checks every question and answer passes through, so the companion only says what a found source supports, keeps other customers out, and masks personal details.

5. 5

   Show the briefing and its sources

   The rep asks “Prepare me for my call with Acme Corp” and gets a briefing built only from Acme's records and the shared ones, with the source of each fact next to it. The app is where the earlier lessons become something a rep can see and check.

6. 6

   Put it where reps work, and watch it

   A companion in a browser tab is one reps forget. So you also put it in Slack, where they already are, and expose it over MCP so other AI tools can use it too. All three call the same backend. Then you watch it in use: is it healthy, is it fast, and are its answers still backed by sources?

How this certificate is scored

Your certificate shows how many of your written answers met the rubric below, checked by an AI grader. Anyone who checks your certificate sees it.

• You answer 6 short questions, one for each module, in your own words (at least 40 words each).
• An AI grader checks each answer against what its module asks for: it must answer the question, use the course's practice case specifically, and be correct.
• To pass, at least 5 of the 6 answers must meet that bar. The certificate says the answers were graded by AI.

The rubric

1. Module 1, Data Lake: Multi-format ingestion pipeline.
2. Module 2, Encode: Chunking + pgvector embeddings.
3. Module 3, Retrieve: Hybrid search with reranking.
4. Module 4, AI Gateway: Prompt routing + guardrails.
5. Module 5, App: Chat UI with citations.
6. Module 6, Deploy: Slack bot + MCP server.

What you walk away with