What a City firm has just said about how lawyers get made
On 9 September, Simmons & Simmons published an argument that legal work will likely shift away from “models built primarily around individual expertise and professional hierarchies” and towards structured workflows, staffed by lawyers working with “legal engineers, data specialists and workflow designers”. Five days earlier, the same firm had published the sentence that actually carries the weight. In a companion piece on legal talent, it wrote that “Not every aspect of professional development will be recoverable through client work” and that some of it “may need to be recognised as a strategic investment in capability-building, similar to research and development.”
Strategic Insight: Those two publications describe one change from opposite ends. The first redesigns how legal work gets delivered. The second concedes that the delivery model which used to train people no longer does it for free.
Why the hierarchy was never only a profit structure
The structure under discussion did two separate jobs at once, and the same invoice paid for both. Junior lawyers produced billable output, which generated margin. That output, as the Simmons & Simmons talent piece puts it, was also the training. “For generations, lawyers have learned by doing,” it says, and the drafting, review, research and analysis juniors once cut their teeth on “were not just tasks to be completed. They were part of the profession’s training ground.”
The firm’s own framing is that these tasks “may have seemed routine, but they were rarely just administrative”, because they taught issue-spotting, argument-testing and where legal risk sits. Then it states plainly that many of them “can now be completed, at least in part, by supervised AI tools.”
The half of the model that has no obvious payer
Strip the first job out of the pyramid and the second does not simply carry on. This is the part that firms have not costed. Delivery redesign has a return you can point at: faster matters, more consistent output, a knowledge base that compounds. Development redesign, on the current framing, is a cost line with a long payback and no client to bill it to.
Simmons & Simmons reaches for the right comparison and then stops. The talent piece says that junior lawyers “may learn more through guided participation in complex work, with experienced lawyers explaining not only what decision was made, but why it was made”, and it notes that “That kind of coaching takes time”. Naming R&D as the analogy sets the accounting treatment, and the piece is clear that the organisation is the one expected to invest. It does not go to the level of that spend, nor to how a firm sustains the line whilst competing on price against one that defers it.
| The number | What it is | Why it binds the redesign |
|---|---|---|
| Two years | The experience requirement on the SQE qualification route, full-time or the equivalent | Whatever replaces AI-absorbed junior work has to fill a two-year window, not a training module |
| Two competences | The minimum a role must expose a candidate to, drawn from the SRA competence statement | A role thinned down to checking AI output may struggle to evidence breadth |
| Four organisations | Maximum across which qualifying work experience can be accumulated | Candidates can spread the requirement; firms cannot subcontract the confirmation |
| 1 May 2026 | Date the Simmons & Simmons bonus recognition scheme took effect | The incentive layer is already moving. The development-funding layer is not |
Strategic Reality: The Solicitors Regulation Authority requires two years’ full-time (or equivalent) qualifying work experience, and that experience has to be the provision of legal services. It must be real, not simulated, which means it still has to come from live matters. Live matters are exactly where AI is absorbing the junior work.
What the redesign actually proposes
From producing documents to running systems
The September piece is specific about the role change. Lawyers are cast as “orchestrators of complex systems rather than sole producers of documents”, working with specialists in workflow, product design and data. It argues that the newer roles are “not supporting functions operating at the periphery of legal services; they are becoming part of the delivery model itself.”
It also identifies the skill that matters most in that arrangement, and it is not tool fluency. The question the firm poses is not how to use AI but “how do we decide when to use AI, and for what?” Distinguishing tasks AI can safely accelerate from tasks that need human judgement is, in its words, “a core professional skill in its own right.”
Critical Context: Knowing which tasks to hand to a tool and which to keep is unusually hard to teach through exposure, because the lesson only lands when someone has felt the cost of getting it wrong. That is precisely the kind of learning the old apprenticeship delivered by accident.
What the piece gets right that most firm commentary misses
- It treats structure as the variable, not tooling. The argument is about how teams are organised, not which platform wins, which is the harder and more useful question.
- It names the incentive problem directly. Traditional models rewarded “utilisation, hours billed and volume of output”, and the piece argues that such measures track delivered value less well once AI compresses the time a task takes.
- It backs the point with its own practice. The firm cites its own bonus scheme, which recognises “meaningful” use of generative AI. Reporting by Tom Cox for Legal Business in June said the scheme had been live since 1 May and “widens the range of non-billable activities that can count towards bonus recognition beyond traditional billable hours”.
- It keeps judgement in the loop. Human judgement is treated as essential where matters are ambiguous or commercially delicate, rather than as a residual category.
Where the argument stops short
The incentive change and the delivery change are both things a firm can decide on its own authority and implement inside a financial year. The development change is not. It runs into a regulator, a qualification route, a client base that has to accept a different pricing conversation, and a competitor set with every reason to let someone else pay for the training and then hire the output.
Hidden Cost: The cheapest strategy available to any individual firm is to let rivals fund the new development model and then recruit the results at three or four years’ post-qualification experience. If enough firms reason that way, the profession underproduces the seniors it needs, and nobody’s board sees the shortfall until it has already happened.
The constraint nobody has costed
What the regulator actually requires
The SRA’s qualifying work experience rules are more specific than the redesign conversation tends to assume. The experience has to be the provision of legal services. The regulator states that “Simulated legal services also do not count”, and requires real life experience instead. A candidate must be exposed to two or more of the competences in the Statement of Solicitor Competence, spread across four organisations at most, and signed off by a solicitor or by the firm’s Compliance Officer for Legal Practice.
Read that against the talent piece’s prescription. Guided participation in complex matters, with a senior lawyer narrating the reasoning, plausibly satisfies the rules, because it is still real legal services. Structured training programmes, experimentation time and coaching sessions do not, on their own. The firm proposes to fund development like R&D, but the qualification route will only recognise the portion of it that remains client work. The gap between what develops a lawyer and what the regulator will count towards qualification is the unbudgeted item.
| Stakeholder group | Primary impact | What they need | How to tell it is working |
|---|---|---|---|
| Trainees and junior associates | The tasks that built judgement are absorbed or thinned; the qualification clock keeps running | Deliberate placement on real complex matters, not just review of AI output | Competence coverage evidenced across live matters, not hours logged |
| Supervising partners and COLPs | Coaching time rises while billable time is the thing being measured | Explicit budget and recognition for supervision, and confirmation criteria they can defend | Supervision hours appear as a funded line, not as goodwill |
| Legal engineers, data specialists, workflow designers | Central to delivery, on the firm’s own account | A progression route, which the September piece does not describe | A published career path that does not terminate at senior manager |
| Clients and in-house buyers | Pay less for output, and may be asked to fund development differently | Transparency about how AI was used and what they are buying | Pricing conversations that survive a procurement review |
What actually separates the firms that manage this
The firms that come through this will be the ones that decided early what a junior lawyer is for, and then staffed and priced accordingly. The ones that struggle will be the ones that cut intake because the economics looked worse in the short run, and discovered four years later that their mid-level bench is thin in exactly the practice areas where clients pay most for judgement.
The second group will not realise they have made a decision. Reducing trainee intake reads as prudence in a year when AI has cut the work those trainees used to do. The consequence lands well outside the planning horizon that produced it.
🎯 Success Factor: Treat trainee and junior headcount as a capability decision with a four-to-six-year payback, not as a cost line responding to this year’s utilisation. Anything else outsources the profession’s supply of seniors to whoever decides to keep paying.
What to do about it, in what order
💡 Implementation Framework: Sequencing the training redesign
Phase 1: Measure what you have already lost (Weeks 1 to 6)
- Audit which tasks juniors did two years ago that AI now handles, by practice area
- Map those tasks to the competences your qualification route depends on
- Identify which competences currently have no reliable route to evidence
Phase 2: Rebuild the route, not the syllabus (Months 2 to 6)
- Place juniors on complex matters in a participating role, not a checking role
- Fund the supervision time explicitly, and say where the money comes from
- Confirm with your COLP what will and will not be defensible as qualifying work experience
Phase 3: Align the incentives with the behaviour (Months 6 to 18)
- Recognise supervision and knowledge contribution in reward, as the incentive redesign already does for technology use
- Test whether your development spend survives a bad quarter, because that is when it gets cut
- Report capability coverage to the board alongside utilisation
Priority actions by where you are starting
If you have not begun
- Name the two jobs separately. Distinguish, in writing, the work your juniors do for margin from the work they do to become capable. Until those are separate lines, nobody can see which one AI removed.
- Ask your COLP the qualification question now. Whether your current junior workload still evidences the competences you assume it does is a question with a factual answer, and it is cheaper to have early.
- Resist the intake cut for one cycle. The saving is real and immediate. The cost is real and delayed, which is the only reason the trade looks attractive.
If the redesign is already under way
- Fund supervision before you fund tooling. The scarce input in the model the firm describes is senior attention, not software licences.
- Build the progression route for non-lawyer roles. If legal engineers and workflow designers sit inside delivery rather than beside it, they need a ladder. A role that is central to delivery and capped in progression will not retain people.
- Price the development line into client conversations deliberately. Clients will notice that matters take fewer junior hours. Decide what you are telling them before procurement asks.
If you are further ahead than most
- Check what your incentive scheme is actually rewarding. A scheme that recognises “meaningful” use of generative AI measures adoption. The skill the September piece identifies as core is deciding when to use it and for what, and adoption metrics only see one side of that decision.
- Instrument capability, not activity. Competence coverage per junior, per practice area, reviewed twice a year, is a harder number to produce and a more honest one than training hours delivered.
- Say publicly what you are funding. In a market where AI tooling is table stakes, a credible development commitment is one of the few recruitment claims a rival cannot match by buying something.
Resource Reality: Phase 1 is a few weeks of partner and knowledge-team time. Phase 2 is the expensive one, because it converts senior hours from billable to developmental, and a mid-sized firm should expect that to be visible in its numbers. That is the point. A development model that does not show up in the accounts is not being funded.
Four things this breaks that nobody budgets for
Evidencing competence when the first pass is automated
If AI produces the first draft and the junior reviews it, the junior may still be providing legal services, but the competences being exercised narrow. Review and verification are genuine skills. They are not the same set as building an argument from nothing, and a qualification route confirmed on the narrower set produces a narrower lawyer.
Mitigation: Have the COLP and the training principal agree, in advance, which matter roles evidence which competences under the new workflow, and staff to fill the gaps rather than discovering them at confirmation.
The funding asymmetry between firms
Development funded as R&D is a cost your competitor can decline to carry whilst hiring the people you developed. The firm that invests looks less profitable for several years and more capable afterwards, and only one of those shows up in the intervening partner conversations.
Mitigation: Commit to the spend at governance level with a stated horizon, so that it is a policy rather than a discretionary line, and report capability alongside profitability so the return is visible before it matures.
Incentives that reward usage rather than judgement
Rewarding “meaningful” use of generative AI is a reasonable way to shift behaviour, and it is a measure of adoption. The skill the September piece names as core professional capability is deciding when AI should be used and for what, and a scheme calibrated on usage rewards only one of the two possible answers.
Mitigation: Pair adoption recognition with a review mechanism that examines decisions not to use AI on sensitive matters, and treat a well-reasoned refusal as a creditable contribution rather than an absence of one.
A delivery model with one career ladder
The September piece places workflow designers, data specialists and legal engineering roles inside the delivery model rather than at its edge. It does not describe how those roles progress, what they are paid relative to fee-earners, or whether they reach the firm’s ownership structure. A model that depends on people it cannot promote is a retention problem waiting to present as a delivery problem.
Mitigation: Publish the progression route for technical roles before recruiting heavily into them, and be explicit about the ceiling if there is one, because candidates work it out either way.
Reality Check: None of this is a one-year programme. A cohort developed under a redesigned route needs two years of qualifying experience before it even qualifies, and several more before anyone can judge its mid-level lawyers. By our own rough estimate, that puts the first real evidence into the next decade. Every intervening year offers a defensible reason to cut the spend.
What this is really a decision about
The September piece closes by suggesting the defining question is not who has the best technology but “who is best able to organise knowledge, talent and judgement around it.” That is the right question, and it is worth noticing which of the three is hardest. Knowledge can be codified, and the firm makes a good case for doing it. Talent can be recruited. Judgement is the one that has to be grown, on a timescale no procurement cycle accommodates, through work that AI is currently absorbing.
Our analysis of the legal AI build-versus-buy divide earlier this month listed the training pipeline among the things that decision breaks. This is the firm-side account of what replaces it, and it is more honest than most. It is also incomplete in a specific way: it identifies a cost and declines to name a payer.
- Separate the two jobs in your financial model. Margin production and capability production were bundled. AI unbundled them. A model that still treats junior headcount as a single number cannot see which half changed.
- Make supervision a funded activity. The redesign depends on senior time moving from billing to coaching, and time does not move because a strategy document says it should.
- Check the qualification route before you redesign around it. The regulator’s requirements are a hard constraint on what a training redesign can look like, and they are knowable today.
Reframing what good looks like
The metric most firms will reach for is AI adoption: seats deployed, matters touched, hours saved. Those numbers are easy to produce and they measure the half of the problem that was never difficult. Buying the tools was never the constraint, which is the same firm’s point when it suggests the defining question may not be who has the best technology.
A more useful board measure is whether the firm can still evidence, for each practice area, that its three-year-qualified lawyers have exercised the competences the work will demand of them at eight years. That number is uncomfortable to produce, which is the main argument for producing it.
Your next steps
Immediate actions (this week)
- Ask your COLP or training principal which competences your current junior workload still evidences
- List the tasks AI now handles that juniors did two years ago, by practice area
- Establish whether any development spend in this year’s budget is protected from in-year cuts
Strategic priorities (this quarter)
- Separate margin production from capability production in the junior headcount model
- Agree, in writing, what qualifies as evidenced experience under the new workflow
- Decide whether supervision time will be funded, recognised in reward, or both
Long-term considerations (this year)
- Set a multi-year development commitment at governance level, with a stated horizon
- Build and publish a progression route for legal engineering and data roles
- Report capability coverage to the board alongside utilisation and realisation
Source: Beyond productivity: reimagining legal teams for the AI era (Simmons & Simmons, 9 September 2026), with its companion Legal talent 2.0: developing skills for the age of AI (Simmons & Simmons, 4 September 2026). Qualification requirements from the Solicitors Regulation Authority’s qualifying work experience guidance. Bonus scheme details reported by Legal Business (reported by Tom Cox, June 2026).
This strategic analysis was written by Resultsense, a UK-focused AI news and analysis publication. We will be watching whether any UK firm publishes a development budget it is prepared to defend in a bad quarter, because that is the point at which this stops being a position paper. Read more analysis at Insights, or get in touch.