Theiagen × Galang AI updates

July 2026 · Monday update

Here's what
we built

One form. Two finished artifacts.

Deliverable 1 of 3 · SOW generator

Step 1 — What ships

Step 2 — How it works

Step 3 — Under questions

What it delivers

One intake form. A SOW and a Quote out.

A single form produces a review-ready SOW document and a Quote spreadsheet, dropped into the client's own Google environment with a human-review checklist.

Review-ready, not a draft to rebuild
Lands in the client's Google Drive
Ships with a human-review checklist

Shipped

Form + testing The intake form and its testing are done and working end to end. Done

v2 · in progress

Drive folders + Notion write-back Output now lands in a per-contract Google Drive folder, then writes the Quote and SOW links back onto the matching Notion contract — matched by the TGN-#### entered as the Quote #. Wired into automation A7, the quote-stage contract scaffold. In progress

Step 1 — What ships

Step 2 — How it works

Step 3 — Under questions

How it works

Template fill, not a language model.

Deterministic template assembly. Form inputs drop into stored templates and write to Drive — predictable, reviewable, the same result every run.

Quote, internal cost tab, and SOW narrative, from the NJ-PHEL SOW template
Pricing comes from the service rates in the stored Quote template
Idempotent — re-running produces the same output

Intake form

Template fill

Drive:
SOW + Quote

Notion
write-back

✓ The Notion write-back runs last and inside a try/catch — a Notion hiccup can never break SOW or Quote generation.

Step 1 — What ships

Step 2 — How it works

Step 3 — Under questions

Under questions

Built to defend.

The questions this tends to draw, and the short answers.

Is this AI or hard-coded rules?

Deterministic template fill from stored rates and templates — not generative. Predictable and reviewable.

How is pricing set?

From the service rates in the stored Quote template.

What if the Notion step fails?

The write-back runs last and inside a try/catch. The SOW and Quote still generate.

Quote-to-Cash Automations

Deliverable 2 of 3 · Section coming from Jas

Where we are

What's next

Open questions

Quote-to-Cash · Where we're at

The whole path works now.

Lead intake through closed won, rebuilt and tested end to end across Twenty, Notion, Slack, and the demo intake form.

Contacts, Qualified Leads and Client Accounts fully covered
Contracts partial: page and status auto, commercial detail by hand
Some runs sit in a secured sandbox while testing continues

Intake

Form submit to Lead A1 to A3: person and company created in Twenty, contact written to Notion, and the deal opened at the Lead stage. Running

Qualified Lead

Discovery stage A4 to A5: on entering Discovery, the client account and lead are created in Notion, with missing-field checks on both. Running

Quote and Closed Won

Quoting, Procurement, Closed Won A6 to A11 plus the SOW Generator: contract created, Slack pings, and status stamped through each closing stage. Running

Where we are

What's next

Open questions

Quote-to-Cash · What's next

Next: the after-signing stage.

Active contracts, renewals, and delivery and invoice monitoring, B1 to B7. The code and structure are already in place; what remains is testing against real cases.

Deliberately the slow part of the work
Most fixes this month came from running real records, not sandbox tests
Twenty edits sync, deletes archive safely, renewals open on schedule

In progress

After-signing, B1 to B7 Invoice schedule, delivery health, data quality, and status derived from dates. Testing

Written

Data migration rulebook First pass done: import with automations off, an LLM to format, a human to check, then stamp the Twenty link. v1

The one rule

Twenty to Notion Once live, data flows one way and never back. Locked

Where we are

What's next

Open questions

Quote-to-Cash · Open questions

A few decisions for Theiagen.

The calls we need from your side to switch fully over from the demo setup.

Which intake form or LLM will be used?

A demo Google Form is running for now.

Which Qualified Lead stages?

Adopt Discovery, Quoting, Procurement, Closed Won, Closed Lost, or keep Theiagen's version.

Who owns a contract once signed?

And the exact handover checklist before it moves on.

Default contract manager and project tier?

Is it Sarah, and how is project tier decided?

AI help for the whole team, the moment they need it.

July 2026 · AI enablement hub

Step 1 — The Hub

Step 2 — Proof it works

Step 3 — What's next

What's delivered

A hub the team actually uses.

A browsable knowledge base for non-engineers, general-first so it works whether someone is on ChatGPT or Claude.

Grounded in the official OpenAI and Anthropic guides
Written for the moment, not to sit unread
AI as quality-of-life, not automation that replaces anyone

30Pages, live

6Areas covered

01FoundationsWhat AI is good at, and what to expect.

02PromptingThe four levers, being specific, giving it material.

03Building blocksSkills, connectors, agents, projects.

04Responsible useWhat's safe to share, always reviewing.

05Getting startedA practical first week, where to get help.

06In practiceSpotting what's worth automating.

Step 1 — The Hub

Step 2 — Proof it works

Step 3 — What's next

The signal it works

It answers real questions.

These came straight from the pilot's actual workflow. The hub handles them today, safely.

Which of my tasks are even worth using AI for?

Watch your week for the repetitive, slow, copy-paste-heavy tasks. Start with the frequent ones, where the payback is quickest.

Can AI help with contract review, and is it safe?

A strong fit for flagging risks. But check the agreement and data policy first, redact or use a sample, and route to a human if unsure.

I redo the same setup every time. Can I save that?

Yes. Do it once end to end, then save it as a skill and reuse it. No re-explaining each time.

Step 1 — The Hub

Step 2 — Proof it works

Step 3 — What's next

How it works

Grounded now, and it stays reliable.

Every answer comes from the maintained hub. It won't invent Theiagen specifics; if it can't verify something, it says so and routes to a human.

Answers naturally, but only from what's in the hub
Built on the real stack, not a generic FAQ
Using it is what keeps it current

Maintained docs

Agent answers

Checked
AI + human

Docs improve

↻ Corrections feed back into the docs, capped so it can't spin. A fix in one place can't silently break another.

Step 1 — The Hub

Step 2 — Proof it works

Step 3 — What's next

Where this is, honestly

Delivered now. Agent next.

Phase 1 is live. Phase 2 is in progress with a real baseline, shown as progress, not a promise.

Phase 1

The knowledge base Thirty pages across six areas, grounded in the official guides, answering the pilot's real questions. Live now

Phase 2

The agent, on top of the hub The structure is built and the review step grades against the full playbook. Early, but a real baseline. MVP end of August

Next: put it in the team's hands and keep it current.

1

Real questions in

Fold the team's actual questions into the hub
Grow coverage where the questions cluster

2

Agent MVP

Reach a self-serve agent that answers the golden-path questions accurately and safely, by end of August.

3

Owner + cadence

Agree who maintains it going forward
Set an update rhythm with you