Article 4.2 · Litigation Budgeting in the AI Age: From Precedent H to Real-Time Dashboards

T3PS Legal Dynamics · Series 4: Practical AI for Legal Operations · Refreshed June 2026

Published as part of the "Practical AI for Legal Operations" series

It is 5:30 on a Friday afternoon. Your phone buzzes. It is Counsel. The CCMC is Monday morning. The costs budget — the Precedent H that should have been filed 21 days before the hearing — has not come in from the other side's solicitors, their own fee earner has not been sent their figures, and someone has helpfully forwarded a version of the form that appears to have been scanned from a photocopy of a fax. You have recommended to try and resolve.

If you have spent any time in litigation management, you will recognise this scene. It happens constantly. Not because lawyers are irresponsible, but because the entire infrastructure around litigation budgeting has been built to guarantee the maximum possible friction at the worst possible moment.

I have been thinking about how to fix this for a long time. The good news is that the tools now exist. The bad news is that the profession has not yet decided to use them properly.

The Precedent H Problem

Precedent H was designed to be simple. It is an Excel spreadsheet. It breaks litigation costs into phases — pre-action, issue/statements of case, CMC, disclosure, witness statements, expert reports, PTR, trial, and ADR — and asks solicitors to estimate their incurred and budgeted costs in each. The court then manages the litigation to those figures, and a successful party's recoverable costs are broadly capped at the budgeted amounts, with good reason needing to be shown for any significant departure.

In principle, this is excellent project management. Costs are brought under control at an early stage. The court has visibility of the financial shape of the litigation. Clients know roughly what they are in for.

In practice, it is a disaster.

The problem is structural. Lawyers do not run their practices by Phase/Task/Activity. They run them by file number, by fee earner, and by time entry. The information needed to complete a Precedent H accurately sits in three or four different systems — the billing system, the DMS, the case management system, and Counsel's head — and none of those systems talk to each other in a way that makes the budget easy to compile. So the budget becomes a reconciliation exercise rather than a planning exercise. It looks backwards rather than forwards. And it is completed under time pressure, by someone who would rather be doing literally anything else.

The email ping-pong that follows is its own form of institutional madness. The budget goes to Counsel. Counsel's clerk emails back asking for an extension. The figures come back with adjustments. The adjusting solicitor emails to query the adjustments. A third version emerges. The judge at the CCMC reviews a badly marked-up PDF and makes annotations in the margins that have to be incorporated manually. At no point in this process is anyone looking at the same document at the same time.

I have spent years watching costs consultants wade through months of emails and WhatsApp messages trying to reconstruct what actually happened, what was agreed, and what was allocated to which phase. The standard answer, which I have given and received many times, is: if it is not allocated, it is not going in. Good work gets written off because the process is broken.

What the Technology Now Allows

The AI is not, in itself, the solution to Precedent H. The AI is the enabler of a better system, and the better system requires a change in process thinking before it requires a change in technology.

But let me describe what I believe that better system looks like, because the components are all available.

The core of it is a cloud-based, self-calculating litigation budget dashboard. Not a static Excel spreadsheet emailed between parties. A shared, real-time environment in which all stakeholders — solicitors, Counsel, client, and ultimately the court — can see the current state of the budget, who is responsible for which tasks, what has been incurred against each phase, and what the projected out-turn looks like against the agreed figures.

This is not a fantasy. Legal project management platforms already offer real-time budget tracking. What is missing is the workflow integration that would make the Precedent H itself a living document rather than a point-in-time snapshot. The Precedent H form, as currently constituted, is not designed for this kind of real-time collaborative use. It is designed to be printed and filed. But the underlying data structure — phases, tasks, incurred costs, estimated future costs — is perfectly amenable to a digital-native dashboard environment.

The implications for the CCMC process are significant. Rather than a specialist judge reviewing a badly scanned PDF with annotations, they could access the live dashboard, review the current figures, add comments and proposed amendments electronically, and issue a costs management order that feeds directly back into the parties' systems. No paper. No transcription errors. No "I thought we agreed phase three at X but the order says Y."

I think of it like the Porsche configurator. You sit down with the client at the outset, you walk through the options — the phases of litigation, the levels of resource at each stage, the contingencies for disclosure complications or expert challenges — and you agree the price. The client understands what they are buying. They can see it update in real time. And the payment-on-account conversation becomes a matter of clicking a button rather than a fraught negotiation six months after the event.

AI's Role in This System

AI does not replace the human judgment required to build a litigation budget. Anyone who tells you otherwise has not tried to budget a multi-party commercial dispute with a difficult expert and an opponent who treats every interlocutory application as an opportunity for satellite litigation.

What AI does is handle the tasks that currently consume enormous amounts of time for no good reason.

Task flow creation. An AI system can take the claim details — the nature of the dispute, the value, the likely complexity, the procedural history — and generate a first-draft task allocation across the Precedent H phases, based on historical data from comparable matters. This does not replace the fee earner's judgment about the specific case, but it provides a structured starting point that takes twenty minutes to produce rather than three hours.

Workflow automation. Budget amendment requests, cost management order responses, phase adjustment notifications — all of these can be automated through workflow tools. The solicitor approves the action; the system executes it, sends the appropriate notifications, and updates the record. The email ping-pong is replaced by a structured, auditable process.

Historical data and machine learning. Here is where the real value accumulates over time. Every budget that a firm files, every bill of costs that goes through detailed assessment, every Counsel fee note that gets agreed or challenged — this is data. Properly captured and structured, it trains a system to understand what litigation at a given complexity level actually costs, how budget estimates from similar firms compare to actual spend, where the consistent variances appear, and what that means for the next budget. The system does not just help you build the budget. It tells you whether the budget is realistic.

Client-facing reporting. The 5:30 Friday crisis exists, in part, because clients have no visibility of the costs position until the solicitor decides to tell them. A real-time dashboard changes this entirely. The client can see, at any point, where they are against budget by phase. Project creep — the slow accumulation of unplanned work that turns a neighbourhood dispute into something resembling the bill for a major infrastructure project — is visible in real time, not discovered at the end of a quarterly billing cycle. The conversation about additional budget happens when it is still manageable, not when it is already too late.

Online Courts and the Productivity Dividend

There is a connected point about online courts and remote hearings that belongs in this conversation. I am acutely aware of the experience of travelling for six hours for a two-hour hearing, working from a train with a VPN that may or may not be secure, arriving at a court building to wait in a corridor before a CMC that takes forty-five minutes and could have been conducted by video call. The bar and the judiciary have been slow to standardise remote hearings, though the Bar Council has called for greater consistency, and there is genuine evidence that remote hearings can reduce hearing times without compromising outcomes.

For a properly designed real-time budget management system, remote and hybrid hearings are not a workaround — they are the natural operating environment. The judge, the parties, and the court administrator can all be looking at the same live document simultaneously. The CCMC becomes a structured review session rather than a performance of document exchange.

The litigation funding market is also moving, if not as fast as expected. The Government confirmed in December 2025 that it intends to reverse PACCAR and regulate the funding market — but the 2026 King's Speech (13 May 2026) carried no funding bill, so the reform is committed in principle and absent in practice, and PACCAR remains live law. Even in that limbo, the direction is clear: an expanded, properly regulated funding market means more claims, more matters, and more pressure to run them efficiently. Real-time budget transparency is not just good practice in that environment. It is a competitive necessity.

The Honest Conversation We Need to Have

None of this works without good data. That is the part of the AI story that gets elided in vendor presentations and conference sessions. You cannot train a system on historical budgets if your historical budgets are filed as PDFs in a matter management system with no structured metadata. You cannot build a real-time dashboard if your time-recording is done in one system, your billing in another, and your matter management in a third, and none of them share a common data model.

The technology is not the hard part. The hard part is the process discipline to capture the right data in the right format from the moment a matter is opened, to allocate costs to phases as they are incurred rather than retrospectively, and to build a culture in which Counsel's fees are agreed in advance rather than presented as a fait accompli two weeks before assessment.

I have been making this argument for a long time. The difference now is that the AI tools exist that would make the system genuinely manageable at scale, even for small and mid-sized firms that could not previously afford the infrastructure. The question is not whether the technology is capable. The question is whether the profession is willing to change the way it works to take advantage of it.

Questions worth sitting with:

1.    If you had real-time visibility of every matter's cost position against budget, how many conversations with clients would you be having differently — and how many problems that currently surface at billing would you catch before they became crises?

2.    Your historical budgets, bills of costs, and fee notes are a potentially significant data asset. Are they currently captured in a format that an AI system could learn from, or are they locked in PDFs and email threads?

3.    When the client asks "how much will this cost?" — how confident are you in your answer, and what would it take to make that confidence justified?