How an answer is reached
Objects first, then reasoning a level at a time
B-Glass never asks a model to read everything at once. It reads each object once (a call, a thread, a ticket, a document), keeps that read, and then builds each level from the level below: roll-ups and insights from the reads, the organisation assessment from those, next steps and the narrative from the assessment, the portfolio from every organisation. There are 12 stages in a declared order.
Why: every sentence is traceable to the records it was made from; the same facts give the same answer on every page; and a change re-opens only the records that depend on it, so the cost stays small.
Figures are never computed by the AI. Renewal dates, price changes, usage trends and reconciliations are computed in code and handed to the model as givens.
What the AI does not read
- Raw exports directly: it reads the entity store and the facts, never the raw files.
- Its own earlier answers about the same subject.
- Anything downstream of the stage it is writing (the assessment never sees the narrative).
- Values that change every day, such as "days ago".
Staticmade from
The stages in order
- object read
- signals
- doc rollup
- person insight
- staff insight
- assessment
- next steps
- narrative
- call prep
- portfolio
- commercial
- tree
Staticmade from
How it fits together
- structure and data
- AI stage
- data flow
- AI reads
- gate
What each box is (18)
- Entity store
- The entity store: one record per thing (225 organisations, 862 people, 1995 calls, 3013 email threads, 1159 tickets, 1218 documents and more), full text, every link held as ids. Every consumer reads it.
- Facts
- Facts: the money and usage views (ARR by month, reconciliation, debtors, usage by month). One source for every number the site and the chat quote.
- Coverage gate
- Coverage gate: every object with text has a read, every organisation has its assessment, next steps and narrative, and the tree is current before the site is built.
- Provenance gate
- Provenance gate: every AI record is checked against its evidence pack (dates, figures and names must be in the evidence; excerpts must be verbatim). Refused records are rewritten inside the loop.
- Grounding brief
- The grounding brief: who BG-Demo is, what it sells and the rules of the world. It is prepended to every AI prompt and its hash is stamped on every AI record.
- Site
- The site: static pages drawn from the tree, readable from disk or any host.
- Assessment
- Assessment: per organisation, a status word with the facts and the opinion behind it, risks and opportunities.
- Call prep
- Call prep: for each booked call in the next 14 days, what to know in one minute, the agenda and the questions to ask.
- Commercial
- Commercial commentary: for each SaaS page of Commercial (dashboard, logos, ARR and deals, profit and loss, ARR, ARR transactions, budget), a short note on what moved, how it reads against the illustrative targets and what to watch. Written only from that page's own measures and the targets file, in layer 5 after the portfolio; a page whose measures are not available yet gets no note.
- Doc rollup
- Document roll-ups: per organisation, what the executed paper says and where drafts and the CRM disagree with it.
- Narrative
- Narrative: per organisation, four paragraphs from commercial position to the closing action.
- Next steps
- Next steps: per organisation, one action with an owner, a customer contact and a date.
- Object read
- Object reads: one AI read per call, email thread, ticket, document and news item, with theme, importance, risks and next steps.
- Person insight
- Person insights: who each person at an organisation is, how much influence they have and what they care about.
- Portfolio
- Portfolio: the Summary blocks across the business: the executive overview and this week (the accounts, the money and the people), renewals, risks, actions, reconciliation and incidents.
- Signals
- Signals: deterministic figures computed in code (renewal dates, price changes, usage trends, who knows the account). The AI is handed these as givens and never computes them.
- Staff insight
- Staff insights: how each BG-Demo employee works across their accounts, including who has left.
- Tree
- The site tree: every page and block, built without a model from the records above, each block carrying what it was made from.
Staticmade from
Records per stage
Updated 9 Oct 26| Stage | Layer | Model | Records |
|---|---|---|---|
| Object read | 1 | sonnet | 7,415 |
| signals | 2 | none (deterministic) | 227 |
| Doc rollup | 2 | opus | 166 |
| Person insight | 2 | sonnet | 862 |
| Staff insight | 2 | sonnet | 58 |
| assessment | 3 | opus | 225 |
| Next steps | 4 | opus | 225 |
| narrative | 4 | opus | 225 |
| Call prep | 4 | opus | 62 |
| portfolio | 5 | opus | 9 |
| commercial | 5 | sonnet | 6 |
| tree | 6 | none (deterministic) |
Computedmade from