Almost every senior employment contract in England contains restrictive covenants. Non-compete clauses, non-solicitation provisions, non-dealing restrictions, confidentiality obligations, garden leave. They are drafted at the point of hiring, when neither party anticipates the circumstances in which they will be tested, and they sit undisturbed in the contract until the relationship ends.

At that point, both parties discover that they have very little idea whether the covenants are enforceable.

An employer who wishes to know whether a departing employee's covenants will hold up in court must instruct solicitors, who will in turn seek counsel's opinion. The cost of that exercise, even at the preliminary stage, is rarely less than several thousand pounds and frequently considerably more. The employee, faced with a letter asserting that the covenants are binding, must obtain equivalent advice in order to understand the position.

In practice, for a substantial number of employees and for many small and medium-sized employers, the cost of obtaining a reliable opinion on enforceability exceeds what they are willing or able to pay.

The consequence is predictable. Employees who cannot afford advice comply with restrictions that may well be unenforceable, forgoing employment opportunities to which they are entitled. Employers who cannot afford to test their covenants either attempt to enforce them without proper analysis (at the risk of incurring costs in proceedings they will lose) or do nothing, allowing restrictions they have paid to draft to go unenforced.

The result, in both cases, is that the quality of decision-making about restrictive covenants is directly correlated with the resources of the party making the decision. That correlation has no principled basis.

A specialist employment barrister, asked to advise on the enforceability of a restrictive covenant, will typically conduct a consultation lasting between thirty minutes and an hour. In that time, counsel will attempt to establish the nature of the restriction, the seniority and role of the employee, the legitimate business interest said to justify the covenant, the duration and geographic scope, the degree of access to confidential information and client relationships, and any circumstances bearing on proportionality.

The quality of that assessment depends, necessarily, on the quality of the probing that takes place during the consultation. The questions counsel asks, the follow-up questions prompted by the answers, and the extent to which all relevant factors are explored within the time available.

The difficulty with the traditional consultation model is not that lawyers are incompetent. It is that the model itself is subject to inherent constraints. Time is limited. The conference typically takes place under fee pressure. Counsel must form a view quickly, and the thoroughness of the questioning depends on the experience, attentiveness, and stamina of the individual barrister on the day.

A barrister conducting a fifth consultation on a Friday afternoon will not, as a matter of human reality, probe with the same systematic rigour as in the first consultation on a Monday morning. The quality of the output varies with the conditions of its production, and neither the client nor the instructing solicitor is well placed to detect that variation.

The enforceability of a restrictive covenant under English law turns on a structured analysis of identifiable factors. The covenant must protect a legitimate business interest. The restriction must go no further than is reasonably necessary to protect that interest. Reasonableness is assessed at the time the covenant was entered into, not at the point of enforcement (as confirmed by the High Court in Patsystems Holdings Ltd v Neilly [2012] EWHC 2609 (QB)).

Each of these factors can be established by asking the right questions, in the right order, with the right follow-ups. The analysis is complex, but it is structured. It admits of systematic treatment.

The legal framework governing restrictive covenants has been substantially stable since the House of Lords' decision in Nordenfelt v Maxim Nordenfelt Guns and Ammunition Co Ltd [1894] AC 535, as developed by Herbert Morris Ltd v Saxelby [1916] and refined by the Supreme Court in Tillman v Egon Zehnder Ltd [2019] UKSC 32, the first decision of that court on restrictive covenants in over a century.

The principles are clear. Their application to particular facts requires judgement, but the framework within which that judgement operates is known and defined. The question is not what the law is, but whether the facts of a given case satisfy the established tests.

Because the legal test is structured, the factual inquiry required to apply it is itself amenable to systematisation. The type of restriction. The duration. The geographic scope. The seniority and role of the employee at the time the covenant was agreed. The nature and degree of access to confidential information. The extent and quality of client relationships. Whether adequate consideration was given. Whether garden leave operates alongside or instead of post-termination restrictions.

Each of these factors can be elicited through a carefully designed sequence of questions, and each answer narrows the range of subsequent questions that are relevant. The process, in other words, is iterative, and it converges on a view.

The development of large language models, and in particular the class of models now described as frontier AI, has produced systems capable of conducting structured analytical work at a level that would have been regarded as implausible even three years ago. These models can read and comprehend substantial bodies of text, apply structured frameworks to open-ended factual scenarios, and generate written analysis of a quality that is, in many domains, comparable to work produced by experienced professionals.

The relevant advantage is not speed, although speed is a consequence. The relevant advantage is consistency. A model applies the same framework, with the same rigour, to every case. It does not tire. It does not abbreviate its analysis because the hour is late or the fee is fixed.

A frontier AI model, given the right instructions, will ask every relevant question every time. It will not forget to ask about the geographic scope of a non-compete because the client gave a long answer about confidential information. It will not fail to probe the circumstances in which the covenant was agreed because the consultation ran over time. It will follow up on ambiguous answers. It will identify gaps in the information provided and ask for clarification before proceeding.

The probing, in short, is as thorough as the instructions permit. And the instructions can be made very thorough indeed, because they are designed by the same specialist barristers who would otherwise be conducting the consultation in person.

RestrictiveCovenant.ai has been designed and trained by ranked employment barristers who specialise in restrictive covenant disputes. The analytical framework, the question sequences, the decision logic, and the system prompts that govern the AI's reasoning have all been developed by practitioners who regularly advise on and litigate these issues.

The model does not replace the judgement of those barristers. It embodies it, in a form that can be applied consistently, at scale, and at a cost that makes specialist analysis accessible to parties who would otherwise go without it.

The user identifies whether they are an employer or an employee. They provide the text of the restrictive covenants under consideration. The system then conducts a structured consultation, moving through the relevant factors in a designed sequence.

At step five, the system departs from a fixed questionnaire and begins to reason about the specific case. It reviews all of the information provided, identifies the areas in which further detail would affect the analysis, and generates targeted follow-up questions. The user answers those questions. The system considers whether the picture is complete. If it is not, it asks further questions.

This iterative process mirrors the structure of a consultation with specialist counsel, with one difference: the system will continue probing for as long as there are gaps in the information, without regard to the time elapsed or the fee incurred.

The system produces a draft advice on enforceability, structured covenant by covenant. For each restriction, it identifies the type, states the legal test, assesses whether a legitimate business interest is established, evaluates proportionality, and delivers a risk assessment.

Following the Supreme Court's decision in Tillman v Egon Zehnder Ltd [2019] UKSC 32, the court's approach to severance has been materially expanded. Where a covenant is found to be too wide, the court may now sever the offending words provided that doing so does not generate any major change in the overall effect of the remaining restrictions.

The system applies this analysis to each covenant under review, identifying not only whether the restriction as drafted is likely to be enforced, but also whether, if it is too wide, the court is likely to save it by severance. That secondary analysis frequently changes the practical outcome, and it requires detailed attention to the precise wording of the clause in question.

The documents produced by RestrictiveCovenant.ai are drafts. They are generated by an AI model, and they should be reviewed by a qualified solicitor before any action is taken in reliance upon them. The system is designed to assist, not to replace, professional legal advice.

The practical effect, however, is that the solicitor who reviews the output receives a structured file: the covenant clauses identified and classified, the factual matrix established, the legal analysis drafted, and the risk assessment completed. The solicitor's task is to review and refine, not to build from scratch. The cost of that review is, accordingly, a fraction of what a conventional instruction would entail.

A conventional opinion on the enforceability of a suite of restrictive covenants, obtained through solicitors instructing specialist counsel, will typically cost between £3,000 and £10,000. The process takes days or weeks, depending on counsel's availability and the volume of material to be reviewed.

RestrictiveCovenant.ai produces a draft analysis in minutes. The cost to the user is a fraction of the conventional fee. The resulting document, reviewed by a solicitor at a low fixed cost, provides substantially the same analytical framework at a total cost that is accessible to individuals and small businesses for whom traditional advice is prohibitively expensive.

RestrictiveCovenant.ai is not a law firm and does not provide legal advice. The analysis it produces is generated by an AI model. It has been designed and tested by specialist barristers, but it can, as all AI systems can, produce errors. The output is a draft, and it is marked as such.

The system does not store user data. The information provided during a session is processed in the user's browser and transmitted to the AI model for analysis. It is not retained after the session concludes.

These limitations are real, and they should be taken seriously. They do not, however, diminish the central proposition: that structured legal analysis of this kind can be made available, reliably and affordably, to parties who presently go without it.