The handover pack, published before you need it

Read the exit clause of a managed service agreement and you will usually find one sentence: the provider will offer reasonable cooperation during transition. It sounds like a commitment. It is a description of an attitude, and an attitude has no table of contents, no delivery date and no way of being incomplete.

The reason that sentence survives is that nobody negotiates it. It sits at the end of a document being signed by somebody who has just decided to hire this firm, and the whole subject is theoretical. It stops being theoretical about three years later, in a two-week stretch when the operations director is trying to work out who holds the account for the site-to-site circuit, and by then every term that decides the answer has been agreed.

So this piece publishes the artifact instead of the attitude. What follows is the handover pack we commit to as a deliverable, the index, what each section answers, and the two design decisions that determine whether any of it is worth having. Take it and hold it against whatever your current agreement says, which is the use we would most like it to have.

The index¶

The index of a handover pack is a directory rather than a document, because the useful version is a thing your next provider can read in an afternoon and work from on day one, not a PDF that has to be re-typed into their own systems.

handover/  00-README.md          what this is, who to call, and what is missing  01-inventory/         devices, accounts, licenses, circuits, contracts  02-identity/          tenants, domains, admin roles, break-glass accounts  03-network/           topology, addressing, firewall rules, circuit details  04-backup/            regime, retention, and the last verified test restore  05-runbooks/          the twenty things that break here, and the fix for each  06-vendors/           who to ring, account numbers, renewal and notice dates  07-open-issues/       what is broken today, and what we were about to do next  08-changes/           what we changed while we ran it, and why  09-credentials.md     where every credential lives, not what it is

What each section is for¶

Sections earn their place by answering a question your next provider would otherwise have to answer by discovery, at your cost, over their first month, while nobody is fixing anything.

Note what section 09 does not say. It records where credentials are held and how custody transfers; it does not contain them. A handover pack full of live secrets is a document that has to be treated as a breach from the moment it is emailed, and it usually is emailed.

The first design decision: it is written continuously¶

The version of this that fails is the one assembled during the notice period. It is written by people who are leaving, against a deadline, from memory, about a system they are no longer motivated to explain, and it is thinnest in exactly the two places the incoming provider most needs it, which are the last two rows of that table.

Sections 07 and 08 cannot be reconstructed at the end. What was broken and what we were about to do about it is a live state that exists on the day and is gone a month later. Why we changed a thing is a rationale, and a rationale written from memory eighteen months after the fact is a guess with a confident tone. Both are only ever accurate if they were written when the work happened, which means the pack is not an exit artifact at all; it is the documentation, kept current, with the exit as the day it gets handed over.

The second: there is nothing to transfer back¶

The pack is the visible half. The half that actually decides what leaving costs is invisible and is settled at onboarding: whose name everything is in. Tenants, domains, license subscriptions and circuits contracted in your name, with administrative access held by you throughout rather than restored on request, mean that a handover is a copy operation. Nothing has to be moved, because nothing was ever somewhere else.

The alternative is the ordinary arrangement in this industry, and it is not usually malice. A provider resells licenses because the margin is real and the procurement is genuinely easier. They register the domain because the client did not have anyone to do it. They hold the firewall under their own management account because that is how their tooling works. None of those decisions is announced as lock-in, and each of them individually is defensible. Together they are the reason changing providers takes six months, and they are why the exit clause is the wrong place to look for the exit cost.

This has a real price for us, and it is worth saying so rather than presenting the arrangement as pure principle. Resold licenses are recurring margin we do not take, and procurement someone here has to do without billing for it. We think that trade is correct for the buyer this division is for, an operations business that has been burned once and is not interested in being locked in twice, and a firm making the opposite trade should at least be made to say which one it is making.

What we do hold, and what happens to it¶

Being honest about this is the part that keeps the rest credible, because “you hold nothing of ours” would not be true. Our monitoring and management agents run on your endpoints, and they are licensed to us. Our ticketing system holds the history of every request your staff have made. Our documentation lives in our systems while we are running the estate.

• Anything we deploy is identified in section 01 while we are still running the estate, including whether it stays useful to you without us. Most of it does not, and saying so early is the point.
• Agents are removed on a schedule you set rather than pulled on the last day, because an endpoint that loses its management agent and gains nothing is worse off than one that keeps it for a couple of weeks.
• Ticket history is exported to you in a readable format. It is the record of what your staff actually struggle with, which is worth more to your next provider than any inventory.
• Documentation is handed over as content, not as a link to a system you will lose access to. A runbook you can only read while you are still a client is not documentation.
• Your data is returned in a platform-neutral format and then certified destroyed, against the NIST SP 800-88 sanitization guidance, with the certificate as a document rather than an assurance.

Four questions for whoever runs your IT today¶

You do not need to be leaving to ask these, and the answers are more useful when you are not. A provider who is comfortable answering them is telling you something; so is one who needs a week to come back to you.

1. 1.Which of our tenants, domains, license subscriptions and circuits are contracted in your name rather than ours? Ask for the list, not the answer in principle.
2. 2.If we gave notice on Monday, what document would we receive, and does a current version of it exist today?
3. 3.When was the last restore actually performed, not the last successful backup job, and what was restored?
4. 4.What have you deployed into our estate that stops working the day you stop being our provider?

None of those is a hostile question and none of them requires you to be unhappy. They are the questions a well-run arrangement can answer in an email, and the ones a badly-run arrangement cannot answer at all, which is exactly the information you wanted before the two weeks in which it matters.

Why publish it at all¶

The obvious objection to putting this on a public page is that it hands a competitor a table of contents. It does, and the table of contents was never the hard part. What is hard is the second design decision (writing it continuously, and holding nothing in our name) and a firm that has not made those two choices cannot produce this pack by copying its index any more than an empty runbook directory makes an estate documented.

The better reason is the one that applies to the reader. An exit artifact that exists only inside a signed agreement is a promise, and you find out what it was worth on the day you use it, once. An exit artifact you can read before you sign is a specification, and you can compare it against three other firms in an afternoon. This division’s whole argument is that the terms which decide what a relationship costs should be legible at the start rather than discovered at the end, and it would be strange to make that argument and then keep the document behind the signature.