Score deals as they arrive
A new opportunity gets scored against your portfolio and every past evaluation, with a first-pass memo drafted and attached to it before the partner meeting.
- Starts
- New opportunity in the CRM
- Steps
- 5
- Checks and gates
- 0
- A person decides
- Never, runs alone
Trigger · Event
New opportunity in the CRM
A new opportunity fires the run the moment it is created, before a partner has opened it.
AttioClaude runs the rest autonomously
- 1
AI confirms the real company.
AttioAttio
The opportunity is tied to a real company by domain rather than by name, and the run stops when that match is uncertain.
- 2
AI compiles the dated public record.
CrunchbaseCrunchbase
Funding, investors, headcount and sector come back with a date on every field, and gaps are recorded as gaps.
- 3
AI scores it against past deals.
The deal is scored on the axes the fund actually decides on, each backed by comparable deals from its own past evaluations.
- 4
AI drafts the first-pass memo.
The memo leads with hard flags, then score and reasoning, the record and open questions, with no recommendation attached.
- 5
AI attaches the memo to the opportunity.
AttioAttio
The memo sits where the partner is already looking, stamped with the date the public record was read.
Every fund has the same bottleneck at the top of the funnel. An opportunity arrives, somebody works out what the company is, pulls whatever is public about it, tries to remember whether the fund has seen something like it before, and forms enough of a view to bring it to a partner meeting. At ten deals a week that is a full day, so the tenth deal gets three minutes.
The inputs are already structured. The public record sits behind an API, and the fund's own history of passes and investments is in the CRM with the reason recorded at the time. Comparing them is mechanical, and it is what gets dropped when the queue is long.
What you get
A first-pass memo waiting on every new opportunity, scored against the fund's own history rather than against a generic rubric, with the reasoning and the sources attached.
What the first pass on a new deal checks
The run fires the moment a new opportunity is created in Attio, before a partner has opened it.
It first ties the opportunity to a real company: legal name, domain, founding year and headquarters, resolved by domain wherever there is one. If the company cannot be resolved with confidence, the run stops and says so, since a memo built on the wrong company is worse than no memo.
It then compiles the public record from Crunchbase: total funding and rounds with their dates, investors on the cap table, headcount and its direction of travel, sector and sub-sector, and any acquisition or shutdown signals. Everything carries a date, because funding data goes stale in both directions and a round that closed eleven months ago reads very differently from one that closed last month. A field that is unavailable is recorded as unavailable. Headcount is never inferred from a funding stage.
Scoring a deal against the fund's own history
The fund's past evaluations come out of the CRM: what was passed on, what was invested in, and the reason stated each time. The new opportunity is scored on the axes those decisions reveal.
- Stage fit, against the cheque sizes and entry points the fund has actually used, which can differ from its stated ones.
- Sector fit, against where the portfolio already concentrates, with high concentration flagged explicitly.
- Conflict with an existing portfolio company, which is a hard flag at the top of the memo and never a component of the score.
- Similarity to past passes. If three comparable companies were passed on for the same stated reason, that reason is the first thing the partner sees.
A benchmark score says the company is a 7 out of 10 and invites nobody to disagree. Scoring against the fund's own past evaluations gives a partner something to argue with. "Comparable to two companies passed on in the last year, both at this stage, both for the same stated reason" hands over the precedent and lets them say why this one is different.
What makes AI deal screening trustworthy
A deal pipeline collects company names as founders typed them into an email, and in a crowded sector two startups with similar names is ordinary. Match on domain instead. Score the wrong one and the memo comes out internally consistent, well sourced, and about somebody else, which nobody catches because nothing in it looks wrong.
Every claim in the memo carries its source and date, and anything unavailable is named as a gap instead of estimated. Open questions are written as questions for the founder, never as assumptions filled in on their behalf.
Withholding a recommendation keeps the memo useful over time. A first- pass document that concludes will, within a few months, be what partners read instead of the evidence, and its errors stop being visible. Hard flags, a score with its workings, and open questions leave the decision where it belongs.
The memo is attached to the opportunity record, where the partner is already looking, in a set order: hard flags, then the score with its reasoning, then the public record, then the open questions. It is stamped with the date it was generated, since the public record it rests on will drift.
The prompt
The real prompt behind the steps above, detailed enough that an agent with the same tools can run it as written. The Run buttons copy the full version, which also carries the steps for connecting those tools.
# Score deals as they arrive
You have access to a CRM and a company data source. Run this on
**[each new opportunity as it is created]**.
## 1. Confirm the real company before scoring anything
Take the opportunity record and resolve it to a real company: legal
name, domain, founding year, headquarters.
Resolve by domain wherever you have one. Company names in a deal
pipeline arrive as whatever the founder typed in an email signature,
and two startups with similar names in adjacent sectors is common
enough that name matching alone will eventually score the wrong
company and nobody will notice.
If the company cannot be resolved with confidence, stop and say so.
A memo built on the wrong company is worse than no memo.
## 2. Compile the dated public record
Gather what is publicly known: total funding and rounds with dates,
investors on the cap table, headcount and its direction of travel,
sector and sub-sector, and any acquisition or shutdown signals.
Note the date on everything. Funding data goes stale in both
directions, and a round that closed eleven months ago reads very
differently from one that closed last month.
Where a field is unavailable, record it as unavailable. Do not infer
headcount from a funding stage.
## 3. Score it against past deals, not a generic rubric
Read the fund's past evaluations out of the CRM: what was passed on,
what was invested in, and the stated reason in each case.
Score the new opportunity on the axes the fund actually decides on,
which are visible in that history rather than in any textbook:
- Stage fit against the fund's actual cheque sizes and entry points,
not its stated ones.
- Sector fit against where the portfolio already concentrates, and
flag concentration explicitly when it is high.
- Conflict with an existing portfolio company. This is a hard flag,
not a score, and it goes at the top of the memo.
- Similarity to past passes. If three comparable companies were passed
on for the same stated reason, that reason is the first thing the
partner should see.
Show the comparable deals behind each score. "Scored low on stage fit,
comparable to two passes at the same stage in the last year" is
arguable. A bare number is not.
## 4. Draft the first-pass memo
Structure it: hard flags first, then the score with its reasoning,
then the public record, then the open questions.
Rules:
- Every claim carries its source and date.
- Anything unavailable is named as a gap, never estimated.
- Open questions are written as questions for the founder, not as
assumptions filled in on their behalf.
- No recommendation. The memo prepares the discussion; it does not
pre-empt it.
## 5. Attach the memo to the opportunity
Attach the memo to the opportunity record, so it is where the partner
is already looking. Include the score, the hard flags, and the date it
was generated, since the public record it rests on will drift.
## Output
Report: whether the company resolved and how, the fields that were
unavailable, the score with its axes, hard flags raised, comparable
past evaluations used, and the memo's location.
Why run it with Tulina
A prompt helps one person. A process moves a team.
Pasted into a chat, the prompt above works once, for whoever pasted it. In Tulina, "Score deals as they arrive" becomes a process the team shares, that keeps its history, runs on its own, and stays inside the same security perimeter as the tools it calls.
The whole team works on the same one
PersonalTeamOrganisation
A process lives where it belongs: yours, your team's or the whole organisation's. Whoever runs it can improve it, and everyone starts from the best version instead of explaining the job again.
Every edit is a version
v1v2v3
Each change is a new version with the previous one kept, so a bad edit is undone by restoring an earlier one. Every run records the version it followed, and what a run teaches goes back into the process.
It runs without you
schedulewebhookon demand
A hosted agent runs the process on Tulina's own infrastructure, on a schedule or when an event arrives, and keeps going when you close the chat. Nothing to host. Every run is tracked to its outcome.
Inside the same security perimeter
vaultrolesrun log
Credentials stay in the vault, never in the prompt. Roles decide who may run what, every call is logged, and a step a person has to approve waits for that person.
Questions about AI deal screening
Related processes
Processes that share tools with this one. The shared tools are lit.
Revive deals that went cold
Deals lost more than 90 days ago come back with fresh signals, contacts checked for job moves, and the features they asked for that have since shipped.
- Shares
- Attio
Monthly schedule6 steps →
Prepare for a customer call
Before the call, the last conversations, the account's state, how the product is actually used, and where requested features stand, in one brief.
- Shares
- Attio
Customer meeting on the calendar5 steps →
