Short notes on running legal work.

When a legal budget bursts — one matter's or the whole department's — the standard explanation is «it got complicated». Sometimes that is true. Far more often something duller happened: the entire job was priced as a single number.

A single number cannot fail partially. It is either right at the end or wrong at the end, and by the time you know there is nothing left to do. An estimate broken into phases fails early and small.

Assumptions on the record

«Fifty hours» is a guess. «Fifty hours, assuming the counterparty is advised, two rounds of comments, and corporate documents in order» is an estimate — and it turns into a negotiating instrument the moment one of those assumptions fails.

Structural optimism

We estimate for the version where everything goes well. The fix is not padding in secret: it is quoting a range instead of a point, and explaining what separates the bottom from the top. Boards and clients both prefer an honest range to a precise number that does not hold.

In practice: break the work into three to six phases, write the assumptions for each, quote a range, and agree up front the variance that triggers a conversation.

A legal department that sends work out without a written scope is signing a blank cheque and hoping for good faith. Usually there is good faith. What there is not is comparability: without a scope, two quotes for the same matter are two numbers that cannot be compared.

The minimum worth writing

It fits on one page: what is being asked, what is excluded, who the point of contact is on each side, what deliverables are expected and by when, and a fee cap with a clear rule for what happens when it is reached — stop and talk, not carry on and invoice.

Keeping the result

Then comes the part almost nobody does: recording the outcome. After a year, a department that logged scope, cost and turnaround for every instruction can say which firm is better at what, with data rather than affection. It is also the only honest way to negotiate rates, because you are finally negotiating over the same thing.

There is an internal effect too. Writing the scope before instructing kills a share of the requests outright — it turns out it was a short opinion the team could do in two hours.

In practice: a one-page template, mandatory above a set value, and a simple log of scope, budget, final cost and a score.

A contract system that only files things is an expensive shared drive. What separates a good implementation is being able to answer five questions without manual work.

One. How long does it take, by contract type, from request to signature? In calendar days, broken down so you can see where the waiting sits.

Two. What share of contracts starts from an approved template? If it is low, either the templates do not work or nobody knows they exist.

Three. How often is a clause negotiated away from the approved position, and who authorised it? Departures are not a problem — they are information. If the same clause is conceded eight times out of ten, it is the approved position that is wrong.

Four. How many contractual obligations have a named owner? An obligation with no owner is one you discover by breaching it.

Five. What share of renewals is decided before the critical date? Renewing by default is still a decision — just one taken by the calendar.

It is rarely one big change that destroys a budget. It is seven small ones: «while you're there, could you look at this annex?», «what if we did it this way instead?». None of them, alone, justifies a conversation about cost or priorities. Together they are worth weeks.

The mistake is treating the record of extra scope as a political act. It is an administrative one: always record, decide later. It can be charged back to the requesting unit, billed to the client, or absorbed knowingly. What it cannot do is disappear — that is how a team ends up with twice the work and the same budget.

Wording matters. «That's outside what we agreed» sounds like a refusal. «I can do that; it adds about six hours and pushes delivery to Thursday — shall I?» is an operational question almost anyone answers naturally.

In practice: an extra-scope field on every matter, filled in when the request arrives, with an estimate and a status. That list is the best material you will have for the conversation about team capacity.

Nearly every assessment we run turns up the same pattern: a substantial share of the work is being done above the level it needs. Legal directors checking formatting, senior counsel compiling bundles, or work sent to an expensive firm when the in-house team would have done it better and faster.

The explanation is never laziness. Delegating means explaining, reviewing, and risking a redo. Under deadline pressure, doing it yourself feels faster — and it is, that time. The problem is that it is always that time.

The cost appears nowhere: the work was delivered and nobody complained. It only becomes visible when you compare who did it with who would have been enough — and that gap often explains the whole budget that was missing.

In practice: when planning, assign each phase the lowest profile capable of doing it well; at close, compare that with who actually did it.

Every legal team has a deadline system. In many of them it is called «Sarah never misses a deadline». That works, right up to the day Sarah is ill, moves on, or has two matters landing in the same week.

A real system has four attributes: a named owner, a backup, notice proportionate to the effort required — warning someone the day before a three-day task is useless — and automatic escalation when nobody confirms.

None of that is sophisticated. It is the difference between depending on exceptional people and having a process that survives ordinary people on a bad day.

The story repeats with almost comic regularity. A system is chosen, training is delivered, everyone starts with enthusiasm. Three months later half the team — and every business unit — is back on email and a spreadsheet.

First cause: people were asked to feed a system that only gave value to management. Whoever fills it in has to get something back.

Second: the old route stayed open. If a request can still arrive by email, it will arrive by email.

Third: everything went live at once. A team learning six modules learns none of them.

In practice: before launch, write down what each type of user gains in the first week. If you cannot write that sentence for the most sceptical person in the building, you are not ready to launch.