Skip to content
SignalPattern
PRIVATE EQUITY/SOFTWARE PORTFOLIOS

Modernize it.
Operate it.
Defend it.

SignalPattern is the AI operating layer for software portfolios: one foundation every portco builds on, one playbook the firm runs everywhere, and a decision file behind every call to the IC.

ONE PLAYBOOK · EVERY PORTCOIllustrative
FIRMYour playbook, as a custom engine
DILIGENCEMONITORINGIC FILES
Portco A
OWN TENANT
ADVANTAGED
Portco B
OWN TENANT
AT RISK
Portco C
OWN TENANT
ADVANTAGED
Portco D
OWN TENANT
RUNNING
Portco E
OWN TENANT
QUEUED
Judgment stays yours. Proof travels.
ONE SYSTEM · THREE JOBS

AI moved the value of every software asset. Your operating model has to move with it.

Same reasoning stack. Same evidence chain. Same governance. Built to build on, from the portco's product to the firm's IC.

MODERNIZE
Build on one foundation?
Chassis · model router · agents · integrations · custom products
01AI-native, not bolt-on
02One base, all portcos
03Partners build
OPERATE · PORTCO
Run it with governed AI?
Pricing · GTM · retention · roadmap · monitoring · execution
01Evidence-backed plays
02Assumptions stress-tested
03Approved work executed
OPERATE · FIRM
One playbook, every portco?
Diligence · monitoring · value creation · IC files · reporting
01Confidence-scored IC files
02Portfolio-wide signals
03Outcomes feed next deal
DEFEND IT · PORTCO RUN

Disrupted or advantaged?

The question every software investor now owes the IC. SignalPattern answers it as a causal graph: typed, weighted drivers, tested against counterfactuals, ending in an action you can sequence.

CAUSAL GRAPH · ILLUSTRATIVE, NOT A RESULT
COUNTERFACTUAL · FORK THE GRAPH
Seat compression from agents
−
Pricing model seat vs usage
±
Switching cost integration depth
+
Data moat proprietary context
+
Adoption pace of customer AI
+
CONFIDENCE-GATEDRetention and growth
VALUE OF THE ASSETAt risk
bar: evidence confidence · + / −: direction of effectBase case: seat compression outweighs the moat. Fork the graph to test a pricing change.
THE QUESTIONIs this product AI-disrupted or AI-advantaged?
THE DRIVERSTyped and weighted
THE TESTCounterfactuals, lenses
THE ACTIONBuild, buy, reprice or retire, in sequence

A better plan says what would change it.

MODERNIZE · BUILD ON IT

Stop bolting AI on. Build on it.

Portcos ship AI-native products on a foundation that is already built: tenancy, auth, audit, privacy, model routing, agents and workflows. Built once, reusable across the portfolio.

THE BASETenancy, auth, audit, privacy. Built.
THE ENGINESRouter · agents · workflows
THE BUILDSpec to working product
THE RESULTAI-native product on your stack, reusable across portcos

We build the base. You build on it.

YOUR SOFTWARE · BUILT BY YOU, ON THE STACK
AI-native moduleNew productCustom engine
SHARED INTELLIGENCE SERVICES · ALREADY BUILT
Model routerAny provider, any tier
Data adaptersFiles in, records out
Signal monitoringWatch, brief, approve
Strategic workflowsDeclared, confidence-gated
Execution enginePlans, tasks, receipts
Belief reasoningMany lenses, dissent kept
Causal reasoningDrivers, counterfactuals
ObservationOutcomes feed back
ENTERPRISE FOUNDATION · ALREADY BUILT
TenancyAuthRolesBillingAdminAuditPrivacyObservability
PORTFOLIO VIEW · SIGNALSIllustrative
Portco ACompetitor moved to usage pricingBRIEF · APPROVED
Portco BSeat count down in two key accountsBRIEF · REVIEW
Portco CDisrupted-or-advantaged run completeIC FILE · v2
Portco DRetention driver re-run on operator inputRUNNING
Portco EPlaybook deployed to new tenantQUEUED
Same engine, each portco. Outcomes from each company sharpen the next run.
OPERATE · THE FIRM

One playbook. Every portco.

Encode the firm's method once as a custom engine. Run it across every company in its own tenant. See the portfolio's signals in one place, and let every outcome feed the next deal.

THE PLAYBOOKYour method, encoded once as a custom engine
THE RUNSame engine, each portco
THE VIEWPortfolio-wide signals
THE LOOPOutcomes from each company sharpen the next run

Every IC file answers the eight questions.

Analysis got cheap. Conviction did not. The decision file is simple enough to sign and structured enough to defend, and it stays inspectable after the deal closes.

01Where did this data come from?
02Can the claim be traced?
03How confident are we?
04What are the weak spots?
05What would change the answer?
06Is the run reproducible?
07What if the model changes?
08Who owns the output?
56,160lens combinations
86production engines
1,100+specialized agents
20thesis families

Open the decision file

Know what supports the decision.

Where did this data come from?

Every source is registered, with run-level lineage from ingest to output.

Can the claim be traced?

Every material claim opens to its source and the transform path behind it.

How confident are we?

Confidence is scored per claim and per task, and weak inputs cap everything built on them.

What are the weak spots?

Evidence gaps are flagged by task, and model disagreement is kept, not averaged.

What would change the answer?

Move the causal drivers to test a different set of assumptions. In TaskExecutor, edit the deliverable to create a counterfactual and revisit the decision.

HOW IT STARTS

Start with one question.

A live decision a portco faces now. Decision file, with receipts. Your operators test it. Expand it or stop. Your call.

Small to start. Proof before scale.
Bring one portco question
THE PILOT · WHO DOES WHAT
YOUR FIRM
SIGNALPATTERN
1 Pick the questionSponsor, portco, NDA first
2 Run itEvidence, drivers, counterfactuals
3 Decision fileSources, weak spots, confidence
4 Test itOperators check it against what they know
5 Re-runOn their corrections, versioned
6 Expand or stopYour call