A bespoke intelligence layer for your firm: intake, document production, meeting intelligence, and follow-through, built on your templates and your data, running on a cloud project you control, owned outright. One build. No per-seat rent.
A private operating system: grounded generation against your document corpus, deterministic verification gates, human sign-off before anything reaches a client. Single tenant. Zero data egress. You hold the keys.
| SYSTEM | private, closed-loop |
| SITE | a cloud project you own |
| TENANCY | single. yours. |
| DATA EGRESS | zero |
| TRAINING ON YOUR DATA | never |
| DRAFT GROUNDING | your templates + matter record |
| FINAL AUTHORITY | a human at your firm |
| BUILD PRICE | 20% of capacity cost, capped |
Every engagement runs on a paper trail. These are the three artifacts you'd hold in your hands, in their real formats. The firm shown is fictional; the formats are exactly what you receive.
rent = seats × tiers × months × every tool that does not talk to the last one. it compounds. never in your favor.
Draft a demand letter for unpaid overtime. Cite controlling authority.
"Pursuant to Smith v. Atlantic Logistics, 412 F.4th 88 (2d Cir. 2019), your client faces immediate liability."
citation fabricated: no such case exists in any reporter"We will pursue treble damages under the FLSA unless payment is received."
wrong remedy: the FLSA provides liquidated damages, not treble"We trust this letter resolves the matter amicably."
no firm template, no matter facts, no venue awareness"Pursuant to 29 U.S.C. § 216(b), your client is liable for the unpaid balance and an equal amount as liquidated damages."
verified: statute retrieved, quotation checked against source textDemand follows the firm's remedy framework for wage claims in this venue.
grounded: framework retrieved from the firm's own knowledge baseRendered in the firm's letterhead template with a provenance appendix.
every claim carries its source; unverifiable claims are flagged, inlineSame prompt. The difference is what the machine is allowed to know, and what it is forbidden to invent.
Delta: retrieval scope, verification gates, template grounding. The model is the smallest part of the system.
Wired to each other and to you. Built from your operating reality in the audit, not from a feature list.
No custom code where a config value will do. Every module below ships with an audit trail, minimum permissions, and one defined job.
Every inquiry qualified, structured, and delivered where it belongs, minutes after it arrives. Nothing re-keyed, nothing dropped.
First drafts generated from your templates, grounded in the matter record, queued for your review with every claim carrying its source.
Every authority checked for existence and quotation before you see it. Deterministic rules decide what needs human eyes, never a vibe.
Consultations summarized, action items extracted, everything filed to the right matter before you're back at your desk.
Your corrections make it sharper. Precedents, playbooks, and templates in a private store that compounds instead of leaking.
Status chased, deadlines watched, and someone accountable for the machine itself, including when a vendor changes an API out from under it. That someone is us, in writing.
| module | contract | mechanism |
|---|---|---|
| intake.route | every inquiry lands where it belongs | poll capture source · match routing table · idempotent |
| draft.grounded | no claim without a source | retrieval-augmented · unverifiable claims bracketed inline |
| verify.gates | authority checked before you see it | deterministic lookup + quote match · no similarity-score shortcuts |
| kb.client_owned | corrections compound, privately | diff capture · dedup · no pooled data, no cross-tenant |
| ops.care | someone accountable, in writing | monitoring · tuning · law-change refreshes · capped care plan |
Your highest-paid people are doing administrative work. That is not a cost you pay a vendor; it is revenue capacity you never invoice. The audit measures it. The formula prices the build against it, capped in writing.
method: recoverable hours × loaded rate × 50 working weeks = annual capacity cost · build = 20% of that number · no hourly meter · no per-seat fees · acceptance criteria written into the blueprint before signing · final payment due only after the system passes them
Billable-practice math: your loaded rate is what an admin hour costs the firm.
Method: recoverable hours × your loaded rate × 50 working weeks = estimated annual capacity cost. We model 75% of mapped admin hours as recoverable; your audit replaces that assumption with a module-by-module figure. Build price = 20% of the capacity cost, capped in writing in the blueprint, with a $12,000 minimum engagement. Scope scales with the price through three tiers. One honest caveat: recovered capacity converts to revenue only when the hours are redeployed into billable or revenue-producing work; the audit maps exactly where those hours go. Estimates, clearly marked; the audit replaces them with your real numbers. Care plan, optional: $750/mo.
Renting line: your estimated annual capacity cost plus the per-seat platform bill, recurring every year, because rented tools store the work while your people keep doing it. Owning line: the one-time build plus the optional care plan at $750/mo, and the per-seat platform gets cancelled. Both lines exclude the 25% of mapped hours we don't model as recoverable; you carry that in either path. Estimates, clearly marked; the audit replaces them with your real numbers.
A focused core: two or three systems against your sharpest bottleneck.
The extended build: four or five systems, wired together.
The full operating system: all six, plus the handover runbook.
The audit's number determines the tier; the blueprint lists exactly which systems are in. Full detail on Pricing.
Everyone else: the audit is free, specific to your firm, and yours to keep either way. Book it below.
Send a few details. We'll map where your capacity is leaking and which systems would win it back first.
You'll work directly with Matthew Samlal, the founder of Synapse Forge. No account managers, no handoffs.
One formula: 20% of the estimated annual capacity cost your audit identifies, with the arithmetic shown line by line. You keep 80% of the first year's recovery and 100% of every year after. The price is capped in writing before work begins, and payment is milestone-based: a signing payment, a mid-build demonstration payment, and the balance at acceptance. Builds carry a $12,000 minimum engagement. Spark ($12,000–$18,000) is a focused core of two or three systems; Forge ($18,000–$25,000) is an extended build; Foundry ($25,000 and up) is the full operating system. The care plan is $750 a month after acceptance, month-to-month. No hourly meter, no per-seat fees. Run your own numbers in the estimator above.
Both are architecture failures, and we architect them out. Every draft is grounded in your templates, your data, and your precedents, tested against your own playbooks and verified authority, and nothing reaches a client until one of your people signs off. Each agent has one defined job, the minimum permissions to do it, and an audit trail of everything it touches. It cannot improvise scope, spend money, or speak for the firm. Every build also ships with a Data Boundary Sheet: a one-page factual summary of where your data lives, what never happens to it, and who signs off, written so your firm can hand it directly to its own clients and carriers when they ask about AI.
The build targets hours you already can't bill: intake re-keying, meeting filing, status chasing, and the re-checking that generic AI output demands. That time comes back as capacity for the work you do bill. Document drafting is the one honest gray area: a recovered drafting hour is only a loss if nobody redeploys it. Firms on flat-fee or volume work capture it directly; strictly hourly firms use it to take on matters they were turning away. The audit maps which of your hours are billable, which are leak, and what recovery is worth at your rates, so you decide with the arithmetic in front of you.
No. The build does not ingest your archives. It runs on a curated launch set: your live templates, current playbooks, and the documents you actually use this quarter. You send those documents as they are, messy is fine; structuring them into the system is our work, priced inside the build, not a project that lands on your office manager. One line we hold: we structure the standards your practice already drafts from, we don't decide them for you. If your partners still disagree about what the template should be, that decision stays with the firm, and the audit will tell you before anything is signed. Assembling that set is a defined step of the build, and the audit tells you exactly what belongs in it before anything is signed. Nobody at your firm is expected to spend a summer indexing old folders, and the system never drafts from a template you've retired. If your templates and intake aren't standardized yet, the audit will say that too, and that finding is free. Before Kickoff you also receive a Pre-Build Readiness Checklist listing every access grant and material the build needs, with time estimates per item; most clients clear the whole list in about an hour.
Because you don't take anything on faith. The audit shows you exactly what we'd build before you commit, the blueprint puts scope, price, and acceptance criteria in writing, and the final payment isn't due until the system passes the criteria you agreed to. The samples in The Work section are the real formats. Judge the work first.
More answers on How It Works and Pricing.