AI joins once the system and your team are ready.

Not before. When it does, it works on the same terms as everyone else: Suggest, Review, Apply. Here's the whole engagement, from audit to hand over, and the mechanism that governs every change we make to your records.

This is the third act. See how we build the system and the adoption that come before it.

Four steps, and you own it at the end.

Every engagement runs the same way. It starts with the audit, so we measure before we build, and it ends with your team holding the keys. AI is the last thing switched on, not the first.

01 · Audit

Measure before we build

Two weeks inside your HubSpot and the tools around it. We score your data quality, process, stack and AI exposure, and hand back a prioritised roadmap with dollar impacts. Fixed fee, credited against a build within 30 days.

See the audit
02 · Build

Build it in your accounts

We clean the data, set the field ownership rules, and wire in the Suggest, Review, Apply mechanism. Every workflow, prompt and piece of documentation lives in your HubSpot from day one, not ours. A build covers Marketing, Sales and Service in parallel, at a fixed price and scope quoted from your audit.

03 · Deploy

Turn it on with your team

We go live with your people, not over a weekend. Reviewers see suggestions in their normal workflow, auto-apply rules switch on one at a time, and we watch the first decisions together until the team trusts it.

04 · Hand over

Keys to your team

Full documentation and training, written assuming we won't be in the room. Keep us on an optional monthly governance retainer, or take it and run it yourself. Either way, you could remove us tomorrow and nothing would break.

Steps two and three run on one mechanism. It's the part that decides whether AI in your CRM is safe. Here's exactly how it works.

Nothing changes without a decision and a record of it.

Most AI tools act first and let you find the mistakes later. We invert that. The AI is allowed to be wrong, because a person stands between every suggestion and your client records.

01 · Suggest

The AI proposes a change

It reads your calls, emails, meeting notes and records, then drafts a specific change: move this deal, update this contact, brief this person. Each suggestion carries its reasoning and its source. A suggestion has no power to act.

02 · Review

A named person decides

The right reviewer, the record owner rather than a central admin, sees the suggestion in their normal workflow. Approve, edit, or reject, in one click. Patterns in rejections feed back into tuning, so the system gets sharper the more your team says no.

03 · Apply

The change lands, logged

Approved changes are applied to HubSpot and written to a permanent log: the change, the approver, the timestamp, the reasoning. Six months later, you can reconstruct exactly why any field says what it says.

Change log · from our own system, anonymised Recent entries
ChangeProposed byDecisionReviewerWhen
Deal stage: Proposal sent → Verbal agreement AI (call transcript) Approved J. Northall 28 Jul, 14:32
Contact role: added "Decision maker" AI (email thread) Approved, edited J. Northall 28 Jul, 14:30
Close date: moved forward to 29 Aug AI (meeting notes) Rejected J. Northall 27 Jul, 09:14
Company industry: standardised to "Legal services" AI (housekeeping rule) Auto-applied Pre-approved rule 27 Jul, 06:00

No field has two writers.

Most CRM chaos has one cause: two systems, or a system and a person, both writing to the same field. So we enforce a single rule everywhere we build. Every field has exactly one writer, named in a register your team holds. If the AI owns a field, people read it. If a person owns it, the AI can only suggest.

Can auto-apply, if you choose

Formatting and standardisation: phone formats, state names, job title casing
Enrichment of empty fields from verified sources, never overwriting an existing value
Internal housekeeping: task creation, list membership, reporting rollups

Each auto-apply rule is written down, numbered, and signed off by you before it runs. You can revoke any rule at any time.

Never auto-applies. Not configurable.

Anything on a client record that changes its meaning: deal stages, amounts, close dates, relationship notes
Outbound communication: nothing leaves in your name without a person approving it
Deletions and merges of any record, ever

This is a line we won't build past, even if asked.

Removable by design.

Some providers run your AI for you and never let go. The system lives in their accounts, the prompts are their IP, and removing them means starting again. We think that's a bad deal dressed up as a service.

We build systems your team owns and runs. Every workflow, integration, prompt and piece of documentation lives in your accounts from day one. We train your people to operate it, not hand them an autonomous agent to babysit, and the documentation is written assuming we won't be in the room.

If you keep us on a monthly retainer afterwards, you're paying for governance and tuning, not custody. The test we hold ourselves to: you could remove us tomorrow and nothing would break.

A governance retainer, if you want one.

The retainer is optional, and we expect most firms to run the system themselves once we hand it over. Some will keep us on monthly to keep the governance sharp as the business changes. It is quoted after the build, never before, and it runs month to month. You can stop it whenever you like and keep everything.

What the retainer covers

A monthly review of what the AI suggested, what your team approved, edited or rejected, and where the suggestions can get sharper
New auto-apply rules written, numbered and signed off as your process changes
Tuning the prompts and thresholds so the audit trail stays clean and the suggestions stay trusted
Training new staff and a named person who answers within one business day

What it isn't

Custody. The system stays in your accounts and your team keeps operating it
A lock-in contract. It is month to month, and stopping it breaks nothing
A condition of the build. The handover assumes you never take it

Either way, the audit comes first. We won't quote a build or a retainer until we've measured what you're working with.

We run our own firm on this.

Seicho's own pipeline, contacts and follow-ups run through the same Suggest, Review, Apply system we build for clients. Below is what a suggestion looks like in our queue, with the client name removed.

You shouldn't take governance advice from a firm that doesn't govern its own data. Ask us to walk you through our portal live on the first call. We will.

See what governed AI would find in your CRM.

The Revenue System Audit. Two weeks, $4,500 fixed fee, under four hours of your team's time. If you don't think it was worth the fee, say so at the readout or email us afterwards, and we refund it in full that day. Credited against a build within 30 days.

Book a Revenue System Audit

Every enquiry answered by a person within one business day.