Ways to start
Start small, with a fixed price and a clear place to stop
This page is for leaders who want agentic AI, data, CRM or integration work moving without betting the budget on a guess. There are six ways to work with us. Each has a fixed shape: who it's for, what you'll hold at the end, how long it runs and how it's priced.
You'll see the scope and the price in writing before any work starts. Every step ends at a point where you decide what happens next, with us, with another firm or on your own.
Find your situation
Where to start
Pick the line that sounds most like where you are today. If none fits, write to us and we'll suggest the smallest useful step.
A supplier has sent you a proposal, and you'd like an independent read before you sign.
Start with a Five-Day ReadYou've been asked to do something with AI, and you're not sure where to begin.
Start with a Footing ReviewYou know the first job you'd give an agent, and you'd like to see it work before you commit.
Start with a Proving RunYour data or CRM is holding everything else back.
Start with a Footing Review, then a Phased BuildYou have an agent live, and you'd like it checked before something goes wrong.
Ask for a Footing Review of the live agentYou know what to build, and you're short of hands.
Talk to us about forward deployed engineeringYour systems are live, and you want them kept tested and improving.
The Long Run is built for thatYou'd like private meeting notes with no bot in the call.
Read how to get early access to Recordist
At a glance
-
Five-Day Read
An independent read of a plan or proposal before you commit.
- Length
- Five working days
- Price
- One fixed fee, invoiced when you receive the review
-
Footing Review
Where to start, what's safe and what to fix first.
- Length
- Two weeks
- Price
- One fixed price
-
Proving Run
One agent on your real data, under rules you've agreed.
- Length
- Six weeks, with a stop point after two
- Price
- A fixed price for each of two phases
-
Phased Build
Your foundation or your agent, built and handed over in phases.
- Length
- Four to eight weeks a phase
- Price
- A fixed price per phase
-
Forward deployed engineering
We join your team and ship with your people.
- Length
- Three months at least
- Price
- A fixed monthly fee
-
Long Run
Live agents and integrations kept tested and improving.
- Length
- Three months, then month to month
- Price
- A fixed monthly fee
Whichever you choose
What comes with every engagement
Whichever way you start, you can count on these:
- A written proposal with a fixed price or a fixed monthly fee, agreed before any work starts. Work you buy from us directly is never billed by the hour.
- The people who'll do the work, named in the proposal. We don't bring in anyone else, including subcontractors, without your written agreement.
- Your code, tests and documents kept in your own repositories and systems from the first day.
- Software, licences and AI usage bought in your name, straight from the provider, so you see the real cost. We don't add a margin to them.
- Rules written down first and turned into tests, wherever we build software.
- Accessibility checked with tools and by hand on anything people will use.
- Written updates in plain words: weekly while we're building, and monthly in the Long Run.
- A clear point where you decide what happens next, whether that's with us, with another firm or on your own.
Review a plan
Five-Day Read
An independent read of a plan or proposal before you commit to it. In five working days, you'll know what's sound, what's missing and what to ask before you sign.
- Length
- Five working days
- Price
- One fixed fee
- Who it's for
- Teams about to approve a plan or sign with a supplier. That might be a proposal for an AI agent, a statement of work or a CRM or data migration plan. It could also be an integration design, or an AI tool's terms and pricing.
- What you get
- A short written review: what's sound, what's missing and the risks, in order of importance.
- The questions to put to your supplier, and what to ask them to add, such as acceptance tests, evaluation after launch, a handover and a way out.
- For any AI tool in the plan, a plain reading of what it keeps, what counts as billable and how the cost grows as use grows.
- A call of up to an hour to talk it through with the people making the decision.
- How long it takes
- Five working days from the day we have the documents and a 30-minute briefing with you. Day 1 is the briefing, days 2 to 4 are reading and questions, and day 5 is the review and the call.
- How it's priced
- One fixed fee, agreed before we start. We invoice it when you receive the review.
- How it starts
- Tell us what you're deciding and when you need to decide it. Please hold the documents until we've agreed a safe way to share them, and check that your agreement with the supplier lets you show them to an adviser.
- How it ends
- With the review and the call. If the plan is sound, we'll say so plainly.
- Our independence
- If we read another firm's proposal for you, we won't bid for that same work unless you ask us to. We read contracts for technical and commercial risk, and we're not lawyers, so we'll point out anything your lawyer should look at. Read how we stay independent.
Find your footing
Footing Review
A two-week review of the ground you'd build on: your data, your CRM, the connections between your systems and the rules an agent would need. You'll know where to start, what to fix first and what to leave alone.
- Length
- Two weeks
- Price
- One fixed price
- Who it's for
- Leaders who've been asked to do something with AI and want a clear, safe place to begin. It also suits teams whose data or CRM feels shaky and who'd like to know before they build on it.
- What you get
- A map of the systems and data your goal depends on, who looks after each one and where the copies disagree.
- Up to three places an agent could help, each weighed for value, risk and how ready the data is, with the one we'd start with and why.
- A first draft of the rules for that agent, in plain words your leadership can read and question.
- A list of what to fix first in your data, CRM, connections, access and accessibility, in order, with a rough size for each.
- A plain reading of the AI tools you use or are weighing up: what they keep, what counts as billable and what they'd cost at your volume.
- A readout with your leadership, and a written report that's yours to share with anyone, including another firm.
- How long it takes
- Two weeks from the day the access we've agreed is in place. Week 1 is short interviews and mapping, and week 2 is analysis, draft rules and the readout. We read settings and descriptions of your data, and we don't need copies of it.
- How it's priced
- One fixed price, agreed in writing before we start. We invoice half at the start and half at the readout. If you go on to a Proving Run within 60 days, we take the review's fee off its price.
- How it starts
- Write to us with your goal, the systems involved and who'd use what you build.
- How it ends
- With the readout and the report. If a next step makes sense, we'll send a fixed-price proposal for it. If it doesn't, we'll tell you, and the report still shows you what to do.
- Already have an agent live?
- The same review works on it. We write down the rules it follows today. Then we try to talk it out of them in a test copy, or in the live one if you agree in writing. We score its answers against realistic questions and list the fixes in order.
Prove it on real data
Proving Run
A six-week pilot that puts one agent to work on your real data, under rules you've agreed, and scores it against a measure you set before we start.
- Length
- Six weeks, stop point after two
- Price
- Fixed for each phase
- Who it's for
- Teams with one clear job for an agent, like answering the questions a service desk hears every day, preparing case notes for staff or sorting incoming requests. You'd like proof before you commit to more.
- What you get
- In weeks 1 and 2, the agent's rules, agreed and signed off, with each rule turned into an automated test. Together we set the measure of success and build a set of realistic questions and tasks to score it against. The people who'll work beside the agent help write its rules and test cases, so they shape it from the first week.
- In weeks 3 to 5, a working agent on your real data, in your environment and on the platform you already use. Plain code decides what it can store, send and show, and a person reviews or takes over where the stakes are high.
- In week 6, a scored run against your measure and the results of our attempts to talk the agent out of its rules. Where it's safe, a few real users try it too.
- Everything we made is yours to keep: the rules, the tests, the question set, the code and the notes.
- How long it takes
- Six weeks from the day the access we've agreed is in place, in two phases. The first two weeks end at a stop point, where we decide together whether to go on.
- How it's priced
- A fixed price for each phase, both agreed before we start. If either of us decides at the stop point that it isn't worth going on, you pay only for the first phase. Any platform or AI usage runs in your own accounts, and we estimate it for you before we start.
- See an example
- Our made-up example for a bike-repair chain shows the rules, tests, question set and go or no-go measure a Proving Run leaves you with.
- How it starts
- Tell us the job you have in mind for the agent, who it'll talk to and where the data lives.
- How it ends
- With a go or no-go decision against the measure you set in week 1. If it's a go, we'll propose how to take it live. If it isn't, you'll have a written account of why and what would need to change.
Build it properly
Phased Build
Once you know what to build, we build it in phases. Each phase has its own fixed price and ends with working software in your environment.
- Length
- Four to eight weeks a phase
- Price
- Fixed per phase
- Who it's for
- Teams ready to put the foundation right, such as cleaning and connecting data, redesigning a CRM or linking systems through APIs and events. It's also how a proven agent goes into everyday use.
- What you get
- A written scope for each phase, with what you'll receive and how we'll both know it's done.
- Working software released into your environment at the end of every phase, with its tests passing.
- Rules and tests that run on every change, and accessibility checked with tools and by hand.
- Notes and runbooks kept up to date as we go, so your team can run each piece as soon as it ships.
- The proposal sets out how we fix any defect in what we delivered.
- How long it takes
- Phases of four to eight weeks, planned one or two at a time.
- How it's priced
- A fixed price per phase, invoiced half at the start and half when you accept it. We price the next phase when the current one ends, and we price any change in writing before we make it.
- How it starts
- A build can follow a Footing Review or a Proving Run, or start from a scope you already have. Send it to us, and we'll reply with questions.
- How it ends
- With a handover. Your team runs releases themselves, and you hold the code, the tests, the notes and the runbooks.
Work inside your team
Forward deployed engineering
We join your team, work in your systems and ship with your people until your team can carry the work alone.
- Length
- Three months at least
- Price
- A fixed monthly fee
- Who it's for
- Teams with a clear goal, more work than people and someone who'll own the result.
- What you get
- The people named in your proposal, working the days each week we've agreed, in your tools, your code and your meetings.
- Small, tested releases, early and often.
- A short written update every week on what shipped, what's next and what's in the way.
- A handover planned from the first week.
- How long it takes
- At least three months, with an end date planned in the proposal.
- How it's priced
- A fixed monthly fee for the days a week we agree, invoiced at the start of each month. After the first three months, either of us can end it with 30 days' notice.
- How it starts
- Tell us the goal, your team and the systems involved.
- How it ends
- With your people shipping without us, and a written record of what was built, what was decided and what to watch next.
Keep it healthy
Long Run
Once your agents and integrations are live, we keep them tested and improving, month by month, whoever built them.
- Length
- Three months, then month to month
- Price
- A fixed monthly fee
- Who it's for
- Teams with an agent or integration in daily use who'd rather not hire a full-time specialist to watch it. Models change underneath an agent, your data changes and people ask new things, so what passed its tests at launch can slip.
- What you get each month
- Your question set run again, with a plain report on what got better, what got worse and why.
- Rule tests kept passing through changes to models, prompts, settings and data.
- A review of a sample of real conversations or actions, within the privacy rules you've set.
- A check of your connections and their logs, so failures are found and fixed early.
- Notice of changes in the tools around you, such as a model being retired, and what each one means for you.
- An agreed amount of time for small improvements.
- A reply to any issue within one working day.
- How long it lasts
- A first term of three months, then month to month.
- How it's priced
- A fixed monthly fee for the scope in your proposal. After the first term, either of us can end it with 30 days' notice.
- How it starts
- After a build or an embedded engagement, or on its own for systems you already run. For work we didn't build, we start with a short, fixed-price review so we know what we're looking after.
- How it ends
- When you give notice, we hand back up-to-date runbooks, a list of every access to remove and the open issues in order.
- Our hours
- We work business hours, and we don't offer round-the-clock cover. If your system needs someone on call overnight, we'll help you set that up with your own team.
Before you spend anything
What's free
Finding out whether we're a good fit costs you nothing. Writing to us, a first call of up to 45 minutes and the written proposal are all free. So are our guides: a checklist for switching on an AI agent safely, and what to put in order in your data, CRM and integrations first.
If you write with a question and there's no work in it for us, we'll still reply with honest pointers.
What it feels like
Your first week with us
This is how the first week of a Proving Run, a Phased Build or an embedded engagement goes.
Day 1
We meet the people involved, agree how we'll keep in touch and ask only for the access the work needs.
Days 2 and 3
We watch the work as it happens today, map the systems and data involved and collect the rules people already follow.
Day 4
We draft the rules and the measure of success, and we go through them with you.
Day 5
You get the first weekly update: what we've learned, the draft rules, the risks we see and the plan for next week.
By the end of the week, you'll hold a map of your systems, draft rules, a measure of success and a plan. They're yours to keep, whatever you decide later.
In writing, before anything starts
What every proposal contains
Every proposal is short, written in plain words and priced before any work starts. It covers:
- the problem, in your words
- what we'll do, and what's out of scope
- what you'll receive, and how we'll both know it's done
- who'll do the work, by name
- what we need from your team, and when
- the timeline, with any stop points
- the price or monthly fee, how it's invoiced and the currency
- any software or AI usage you'd buy in your own name, with an estimate of its cost
- how we'll handle your data, and who owns what we build
- how either of us can end it
Honest about limits
When we're probably not the right fit
We'd rather tell you now than halfway through. We're likely the wrong choice if you need any of these:
- someone on call around the clock
- a supplier with a formal security certification, such as SOC 2 or ISO 27001
- dozens of people on a project at once
- an agent live in days, before its rules have been written down and tested
- a firm that resells software licences as part of the deal
If one of these rules us out, write anyway. We'll tell you what to look for in another firm.
Before you decide, read the questions to ask before you hire us, and how we look after your data and systems.
Not sure which fits?
Write a few lines about what you're trying to do. We'll reply within two working days with the smallest step that would tell you something useful, and the earliest date we could start it.
Email hello@havihi.digital Read the questions to ask before you hire us