Pilot programme

A 5-working-day pilot. Named outcomes. Named exit.

The fastest way to know whether HookBus and CRE-AgentProtect Enterprise fit your organisation is to run them against one real workload for one working week. Scoped on Monday, installed by Tuesday, measured by Friday, decision the same day.

Duration · 5 working days, Monday to Friday Scope · one production-shaped workload Team · one engineer, one security lead Exit · defined before we start

Specifics agreed on Monday. Options, today.

Every pilot is shaped around your data-handling posture, your deployment constraints, and what your DPO will sign off inside a working week. Three typical shapes, and we will agree the specifics together on day one.

Shape A

Your estate, your data

HookBus and the subscribers install inside your environment. Real data flows only through a DLP-redacted envelope. No content, PII, secrets, or regulated data leaves your estate at any point.

  • On-prem, VPC, or air-gap
  • Your IAM, your network
  • Nothing exits your perimeter
Shape B

Our hosted environment, synthetic scenarios

HookBus runs on Agentic Thinking infrastructure. Zero PII, zero customer data. We run a set of pre-agreed synthetic scenarios your team signs off before Monday. No DPIA needed to start.

  • Fastest path to seeing it work
  • No customer PII ever touches our servers
  • Scenarios locked before install
Shape C

Hybrid: your bus, our dashboards

HookBus core and subscribers run in your VPC. Observability dashboards are served from a read-only shared surface. You retain data residency; we help you interpret the evidence pack in real time.

  • Data stays in your VPC
  • Shared dashboard view only
  • Our team watches alongside yours

Locked before we start

Whichever shape you pick, the exit criteria and the content scenarios are agreed together on Monday. No surprises Friday. No moving goal posts. If the pilot is a no, the exit is defined before the install.

Monday scope. Friday decision.

Every pilot runs on the same one-week cadence. Every step, every artefact, every outcome is committed before we start.

Monday

Scope & provision

90-minute kickoff. One workload, one publisher, subscribers and success criteria locked. Scoping document signed same day. Environment ready to install.

  • Scoping document signed
  • Success criteria locked
  • Non-binding evaluation MSA signed
  • Deployment target ready
Tuesday

Install & instrument

HookBus and the chosen subscribers go live in your environment. Publisher wired to the pilot workload. Dashboards accessible to your team. Smoke test on a non-sensitive event.

  • HookBus live in your environment
  • Publisher wired to real traffic
  • Auditor and DLP Filter running
  • End-to-end smoke test passes
Wed – Thu

Run & measure

The pilot runs against real work. Daily 15-minute stand-up. Policy iterations in-flight. We tune and you watch. Evidence accumulates in the Auditor chain.

  • Daily stand-up, 15 minutes
  • Red-team list run against L1/L2
  • DLP match counts measured
  • Latency and decision counts captured
Friday

Review & decide

90-minute wrap. Evidence pack delivered. Success criteria scored line by line. Decision same day: proceed to purchase, extend one more week, or exit cleanly.

  • Evidence pack delivered
  • Scoring against criteria shared
  • Commercial proposal ready if go
  • Clean uninstall if no

What we agree before the first install.

Every pilot starts with a short scoping document, completed together on Monday. Specifics are deliberately left open so your team shapes them; the fields below are what needs a decision on day one.

FieldWhat we write in
Workload One named publisher running against one named workload. For example: "Claude Code used by platform engineering team against internal infrastructure repos" or "an autonomous agent calling the treasury-settlement API in the dev environment".
Environment Pre-prod, UAT, or production. On-prem, VPC, or air-gap. We match deployment to what your DPIA allows. Never production without an approved DPIA.
Subscribers enabled Default bundle is Auditor plus DLP Filter plus CRE-AgentProtect L1. You add Session Memory, KB Injector, or CRE-AgentProtect Enterprise L2 if they apply.
Success criteria Three to five measurable outcomes your team sets. "Every agent tool call produces an Auditor row" is a criterion. "We like it" is not. We score against these on Friday.
Data handled What regulated data (PII, PHI, PAN, session tokens, credentials) the pilot workload may touch, and which DLP patterns to enforce. Signed by your DPO.
Security questionnaire SIG Lite, CAIQ, or your bespoke vendor questionnaire. We return a completed response within 48 hours of kickoff. Reviewed by your security team in parallel with install.
Stakeholders One engineering lead, one security or DPO contact, an executive sponsor. We name ours too: the founder runs every pilot personally.
Commercials The pilot is free of charge against a signed non-binding evaluation agreement. Scoping-only MSA. Neither side is committed past Friday.
Exit ramp If you decide no on Friday, we uninstall the following Monday, hand back the Auditor archive, wipe the HookBus volume. No claim on future business. The evidence pack is yours to keep.

What makes a pilot a pass.

Examples. Your criteria will be your own. These show the shape.

Every tool call is logged

The Auditor subscriber produces a row, hash-chained, for every PreToolUse and PostToolUse event on the pilot workload.

Metric: 100 percent of events logged, zero chain breaks

Dangerous commands are blocked

CRE-AgentProtect Enterprise deny-wins on a curated list of dangerous tool calls drawn from your own incident history.

Metric: zero false negatives on red-team list, max 2 percent false positives

Regulated data never leaves the boundary

DLP Filter redacts PII, PAN, and secrets on both tool_input and tool_result before any subscriber persists the event.

Metric: zero regulated-data strings in the Auditor archive after redaction

No outbound calls at decision time

L1 and L2 decisions run locally. Your network-egress telemetry shows zero traffic to Agentic Thinking infrastructure during the pilot.

Metric: zero outbound packets to *.agenticthinking.uk, *.hookbus.com during run-phase

Under-10ms decision latency p95

End-to-end latency from publisher emit to bus verdict for L1 policy decisions stays under 10 milliseconds at p95 under your pilot traffic.

Metric: bus-side latency p95 < 10 ms across the Wed–Thu run period

Your compliance team can read the evidence

The Auditor export is in a format your internal audit or external assessor can actually review. Not a dump. An evidence pack.

Metric: internal audit sign-off on format, within 2 business days of delivery

Pilot fit criteria.

We say no to pilots that are not a fit. Saves both sides time. Here is how we qualify.

Good fit for a pilot

  • Regulated industry: banking, insurance, healthcare, defence, public sector
  • Already running AI agents or coding assistants against production data
  • CISO, DPO, or Head of AI Governance sponsor identified
  • Procurement timeline of 3 to 12 months from pilot end
  • Technical team can spare one engineer for a half-day on Monday and 15 minutes daily Tue–Fri
  • Legal team can turn round a one-page non-binding evaluation agreement inside 48 hours

Not a fit for a pilot

  • Fewer than ten agent or coding-assistant seats in production
  • No named executive sponsor
  • Unwilling to commit an engineer to a Monday kickoff and daily check-ins
  • No clear pathway to a purchasing decision within twelve months
  • Expecting the pilot to be open-ended or exploratory
  • Expecting free perpetual use after the five days

What the pilot costs, and what comes after.

Pilot

Free of charge, against a signed evaluation agreement

The five-day pilot carries no licence fee. You cover your own deployment infrastructure. We cover our time, the software, and the direct support. Non-binding on both sides.

Purchase

Annual commercial licence

If the pilot passes and you choose to proceed, HookBus Enterprise and the licensed subscriber bundle are sold on an annual basis. Typical engagement from £300,000 per year depending on scope.

Extend

Second week, paid

If the Friday decision is inconclusive and both sides want more time, we run a paid second week against a narrower scope with revised success criteria. No further extension.

If you decide no, we leave cleanly.

The whole point of a pilot is that the answer can be no. Here is what no looks like when it happens.

Uninstall the following Monday

If Friday’s decision is no, we uninstall the next working day. HookBus and every subscriber we deployed are removed. Any persistent volumes we created are handed to you or destroyed to your preference.

Evidence pack stays with you

The Auditor archive, the policy packs written for your workload, and the scoring sheet against the success criteria are all yours to keep. You can use them inside your own compliance programme.

No claim on future business

No renewal, no right-of-first-refusal, no contractual follow-on. The scoping agreement terminates cleanly.

We ask for feedback

Thirty minutes with your engineering lead and security sponsor. What broke, what worked, what we should change. We use it to make the next pilot better.

Start a pilot scoping call.

Thirty minutes with the founder to agree whether a pilot makes sense, and what the scope looks like if it does.