Discover the work. Deploy the workforce. Prove the ROI.
ProcessTwin connects to the tools your company already uses, maps how work really flows, prices every automation opportunity in pounds, and runs AI agents you can actually govern — with an auditable savings ledger behind every number.
Free trial · No credit card · Agents stay in suggestion mode until you say otherwise
From “where does the time go?” to a working digital workforce
Four steps, in plain sight. No black box — every number on your dashboard traces back to the evidence underneath it.
- 01
Connect
Plug in Gmail, Slack, Salesforce, QuickBooks and more. Read-only scopes, PII masking on by default.
- 02
Discover
ProcessTwin maps recurring tasks, approval chains and duplicated work into a living process graph — bottlenecks highlighted.
- 03
Score
Every opportunity is priced: impact, difficulty, hours saved, annual pounds, AI confidence. Tuned to your hourly rates.
- 04
Automate
Deploy AI digital workers with autonomy dials, approval gates and a kill switch. They must pass evals before going live.
Everything between “we should automate this” and audited savings
Discovery engine
13 connectors detect recurring tasks, approval chains and duplicated work from the tools you already use.
Process graph
An interactive map of people, systems, documents and approvals — cycle times per step, cost overlays, bottleneck alerts.
Opportunity engine
Composite scoring in pounds with payback periods, tunable weights and a quick-wins portfolio view.
Governed AI agents
LangGraph-powered digital workers with human approval gates, spend caps, tool allowlists and full run traces.
Savings ledger
Every hour and pound saved is logged with evidence, and split verified vs projected. ROI you can defend to a board.
Enterprise security
Per-tenant encryption keys, tamper-evident audit chain, prompt-injection quarantine, UK GDPR tooling — security first.
AI agents your security team will actually approve
ProcessTwin was designed around the OWASP guidance for LLM applications and multi-tenant SaaS. Guardrails live in the execution layer — never just in a prompt.
Read the security overview- Per-tenant encryption keys
- Tamper-evident audit chain
- Prompt-injection quarantine
- Human approval gates + kill switch
- Row-level tenant isolation
- UK GDPR export & deletion built in
Automation you can audit.
Every agent action writes to a savings ledger with evidence. Finance flips entries from projected to verified. Your ROI graph is defensible line by line.
Frequently asked questions
How does ProcessTwin discover our processes?+
You connect the tools your team already works in (email, chat, CRM, finance). ProcessTwin reads the minimum metadata needed, detects recurring patterns — approval chains, handoffs, repeated tasks — and assembles them into an editable process graph with confidence scores and evidence.
Is our data used to train AI models?+
No. Your data is processed to serve you and never used for model training. You can additionally enable PII redaction so personal data is masked before any AI provider sees it, and choose which provider handles your workspace.
What stops an AI agent from doing something it shouldn’t?+
Guardrails are enforced in the executor, not the prompt: per-agent tool allowlists, spend caps, blast-radius limits, human approval gates for external actions, an instant kill switch — and agents must pass an evaluation suite (including adversarial prompt-injection tests) before deploying.
Where does ProcessTwin run?+
ProcessTwin is a fully managed platform — we run the infrastructure on hardened EU servers so you never patch, scale or babysit anything. Every workspace is isolated with tenant-scoped queries, row-level security and per-tenant encryption keys, and Enterprise plans can run on a dedicated single-tenant instance.
How is the ROI number calculated?+
Savings are recorded in a ledger entry-by-entry — what an agent did, hours saved at your configured hourly rates, evidence links. Entries start as “projected” and you (or your finance team) flip them to “verified”. Dashboards always show the split, never a blended guess.
How long does setup take?+
Minutes to connect the first tool. First discovery results typically appear within the hour, and you can deploy your first governed agent within the first week — starting in draft-only mode until you trust it.
Can we start small?+
Yes — the free trial supports one connector and suggestion-mode agents. Autonomy levels mean you can keep humans approving every action for as long as you like.
What happens if we leave?+
Your data is yours: export everything from settings at any time, and workspace deletion is a real deletion — a 30-day grace period, then hard removal including derived data and encryption keys.
Ready to see your processes?
Connect a tool, watch discovery light up, and deploy your first governed agent — free.