Your business process operating system
Hand off the work you dread. Keep the work you love.
tacitrun turns the repetitive, swivel-chair work across your systems — the dropdowns, the dialog-after-dialog, the copy-paste between apps — into AI domain agents your IT will actually approve. So Sales, Revenue, Commerce, and Customer Service Ops can spend their days on customers and growth, not busywork.
Start in minutes: connect a system, pick a blueprint, and watch a domain agent run alongside your team in shadow mode before it touches anything. Every write stays behind your approval until you decide otherwise.
Don’t start from a blank page. Start from your industry.
tacitrun ships your industry’s operations already mapped — every process with its steps, KPIs, suggested systems and human-approval points — and pre-built domain agents ready to run them. Adopt one in a click, then adapt it to your systems, your terminology and your approval rules. Your people review and refine; nobody writes it from scratch.
You run operations, and you’re tired of being the human glue between systems.
You want to get rid of the repetitive tasks you dread doing every week.
You’d rather be talking to your customers than wrangling tools.
You want to spend your time on strategic work that grows the top line.
You swivel-chair across a dozen apps — dropdowns, dialogs, endless screens in your systems of record — and you’re done with it.
Adopt what’s built. Adapt it to how you work.
Start from your industry’s blueprint or a pre-built domain agent, and adapt it to your systems, your terms and your approval rules — in an afternoon, not a project. Something we haven’t built yet? If your team can describe it — typed, spoken, or the SOP you already have — tacitrun can run it. Either way, a person approves before anything changes in your systems.
Your reps should be selling, not scrubbing data. Hand tacitrun the routing rules, territory sheet or hygiene checklist you already follow — it routes, cleans and preps the pipeline, and you approve what changes in the CRM.
Lead Router · CRM Hygiene Fixer · Deal Desk Reviewer and more
- Hi-tech / SaaS: Route inbound leads by territory and fit; flag stale deals before the forecast call.
- Staffing: Turn new client requests into qualified leads, assigned to the right recruiter desk.
- Commercial insurance: Send each broker submission to the right underwriter, reasons attached.
Quote-to-cash breaks in the gaps between systems. Describe your deal-desk policy or reconciliation checklist — tacitrun checks every deal across CRM, quotes, contracts and billing, and routes each exception to the person who owns it.
Forecast Roll-up · Pipeline Hygiene Auditor · Commission Calculator and more
- Hi-tech / SaaS: Check non-standard discounts against policy and route them to the right approver.
- Consumer lending: Catch missing application data before it stalls underwriting.
- Private equity: Keep deal and portfolio data reconciled across CRM and your data warehouse.
Every new SKU, price change and stock update is a chance to get it wrong. Give tacitrun your catalog playbook — it imports, enriches and syncs products across PIM, ERP and storefront, and your merchandiser approves before anything goes live.
Product Listing Enricher · Catalog QA Auditor · Replenishment Planner and more
- Retail: Draft store replenishment orders from stock and sales; approved before the PO goes out.
- Consumer goods: Match retailer deductions to promotions and route the disputes.
- Manufacturing: Keep product specs consistent between ERP and the online catalog.
Your team should be solving problems, not sorting the queue. Hand tacitrun your triage rules, macros and escalation policy — it routes tickets, drafts replies from your own knowledge base and flags what is about to slip, and your team approves what changes.
Ticket Triage · Response Drafter · Escalation Watcher and more
- Hi-tech / SaaS: Triage tickets by product and urgency; draft replies from your help center.
- Retail: Decide refunds within policy and escalate the exceptions to a person.
- Healthcare payer: Route member questions to the right team, with the history attached.
The business sees the hours handed back and the value delivered by every process. IT sees how often each agent agrees with your team and what every run costs. Same numbers, one place — so the case for AI is measured, not assumed.
Not on this list? If your team can write it down, tacitrun can run it — including the back-office work every company runs: Billing Ops · Contract Ops · Finance Ops · HR Ops · Order Management · Procurement.
Business builds. IT approves. One seamless handoff.
The business moves fast without waiting on a backlog; IT keeps control without becoming a bottleneck. Both sides see the same live status for every domain agent — Ready, Pending IT, Needs IT setup, Live — so nobody has to guess where things stand.
Pick a blueprint or describe the work in plain words. tacitrun builds the process and its domain agents; the business checks them safely — no code, no tickets.
When it’s ready, the business submits the whole process for IT review. Every domain agent goes together, as one — no piecemeal requests.
IT connects your enterprise systems, binds the fields, and signs off. Nothing goes live until it’s connected and approved — governance by default, no shadow IT.
Bring your own model. Run agents in your own GCP project.
Large enterprises do not send their conversations through someone else’s API key. On the Enterprise plan the model inference and the agent runtime both live in your cloud — your project, your identity, your bill, your audit trail. tacitrun keeps building the prompts and holding every write for approval; the model and the agents run where your security team wants them.
Your models, in your GCP project
Every model call the platform makes for your workspace — composing, runs, evaluations, chat, reading your documents — runs on the Claude models you enable in Vertex AI Model Garden, inside your own Google Cloud project.
- Keyless, via Workload Identity Federation
- Billed to your project
- In your audit logs, under your IAM
- Inside your VPC Service Controls and CMEK
Or your own OpenAI / Azure OpenAI key
Prefer OpenAI? Bring your own API key or your Azure OpenAI deployment and run on:
- Stored encrypted in the vault
- Used only for your workspace
- Never shown again
- Removed the moment you switch back
Your agents, running in your project
Each deployed domain agent becomes a Cloud Run service created in your GCP project, calling Claude models on your Vertex AI.
- Runs as a service account you name
- Reads its manifest from your bucket
- Public URL that only tacitrun’s identity may invoke — or internal ingress behind your own load balancer
- Nothing but the tool callback leaves your perimeter
Set up by an owner under Settings — Cloud deployment and Bring your own model. See the Enterprise plan · How it works, in the FAQ
A look at the actual product — three views you'll spend time in.
A blueprint library for every industry
Pick your industry and see its processes grouped the way you actually think about them — Front office, Digital commerce, and the Standardized finance / HR / IT / procurement / legal back-office every company runs. Aligned to common process frameworks. One click seeds a governed process pre-filled with steps, KPIs, suggested systems, and human-approval points.
Import from Celonis or SAP Signavio
Already mapped a process in Celonis or SAP Signavio? Import what they know — tacitrun deconstructs it into tasks, flags the repetitive, automatable steps, and recommends the specific domain agents to run them so your people are freed for strategic work. Imported knowledge lands in your corpus and feeds the same compose → eval → shadow → approve → deploy pipeline. Try it with sample data, or connect your own tenant.
Deployments dashboard with live health
Every deployed agent shows real-time liveness, 24h invokes, p95 latency, and 24h cost. Polled live from the agent's own /health endpoint + the invocations we proxy. No leaving tacitrun to dig in your cloud console.
| Agent | Health | 24h | p95 | err |
|---|---|---|---|---|
| PO Approval Domain Agent | ok | 247 | 2.8s | 0% |
| FNOL Triage | ok | 184 | 1.9s | 1% |
| Trade Promo Reviewer | ok | 92 | 3.4s | 0% |
| COI Renewal Scheduler | down | 6 | — | 100% |
Observability built into the runtime
Every invocation persists a trace span with cost, tokens, latency, status, and redacted I/O. Filter by agent. Click into a trace to see the full LLM + tool-call timeline. No separate OTel collector to operate.
| Agent | When | Status | Dur | Cost |
|---|---|---|---|---|
| PO Approval | 12s ago | ok | 2.4s | $0.014 |
| PO Approval | 38s ago | ok | 2.9s | $0.012 |
| FNOL Triage | 52s ago | ok | 1.8s | $0.008 |
| PO Approval | 1m ago | escalated | 3.1s | $0.018 |
| Trade Promo Reviewer | 1m ago | ok | 4.2s | $0.024 |
The full end-to-end stack. Not a demo — a working system.
Upload PDFs, screenshots, or describe in chat. tacitrun extracts the structure into a typed process graph with actors, decision rules, SLAs, and compliance tags.
Cross-industry foundations, end-to-end value chains, and per-industry packs — pick your industry and they group into Front office, Digital commerce, and Standardized back-office. Aligned to common process frameworks. One click seeds a governed process.
Connect Celonis or SAP Signavio (or try sample data) and import a process you've already mapped. tacitrun deconstructs it into tasks, flags the automatable steps, and recommends the domain agents to run them.
The composer reads your corpus AND the graph. Generated agents use your business's exact vocabulary, named approvers, real thresholds — not generic templates.
Every agent gets evals derived from concrete scenarios in your SOP. Judge rubrics embed the source-of-truth excerpts so the grader has context at run time.
Watch the agent on real work alongside your humans. ≥90% agreement promotes it out of HITL. Below that, you keep the human in the loop.
Lock-on-approval governance: agents in production are frozen until an explicit unlock. Promotion requests track 5-dim risk scoring + cost projection.
GCP Cloud Run or Vertex AI Agent Engine, today. tacitrun-hosted by default; on Enterprise, each agent’s Cloud Run service is created in your own GCP project, as your service account, keyless via Workload Identity Federation — plus Gemini Enterprise registration.
Deployed agents are IAM-gated. No public endpoints. All invocations go through a server-side proxy that mints Google-signed credentials per request.
Every invoke writes a structured trace + invocation row. Health dashboard shows p50/p95 latency, 24h cost, error rate — without leaving tacitrun.
Frontier Claude models by default. On Enterprise, run every model call on Claude models on Vertex AI inside your own GCP project — no API key shared — or on your own OpenAI / Azure OpenAI key with GPT-5.6 Sol, GPT-5.6 Terra or GPT-6 Astra.
Pick one process your team dreads. Have it running this week.
Start with a blueprint for your industry, connect the systems it touches, and run it in shadow mode against real work. You see the agreement rate and the exact writes it wants to make before anything is approved.