A practice partner I work with already had Microsoft 365 Copilot rolled out across her firm and was being pitched by Spellbook on the same day. She asked whether she had to choose. Her instinct was that one tool had to be the right answer. Once we walked the use cases through, she saw they were doing different jobs and the right answer was both.
By 2026 the general-purpose AI platforms and the sector-specialist tools have stopped being competitors and started being layers in the same stack. The plain-English version tells you which layers you actually need.
The choice you’re facing
General-purpose AI platforms are designed for breadth. ChatGPT Enterprise, Claude Teams (£20-£100 per seat per month depending on tier), Microsoft 365 Copilot (about £21 per user per month on top of the base Microsoft subscription) and Google’s Gemini for Workspace. They sit inside Outlook, Word, Excel, Teams, Gmail and Docs, and handle generic tasks well: drafting, summarising, document and spreadsheet work, mixed-subject Q&A.
Vertical specialist tools are designed for one sector’s workflow. Harvey for enterprise legal, Spellbook for contract drafting in Word, Eve.legal for plaintiff litigation, Sage Copilot for accountancy month-end close, Dext and FreeAgent AI for smaller practices, Buildr for preconstruction in construction, Bullhorn Amplify and hireEZ for recruitment. They come pre-trained on the sector’s templates, terminology and regulatory framework, and integrate with the case-management, ledger or ATS platforms your practice already runs on.
The defining difference is depth versus breadth. General-purpose AI is configured by you through prompts and retrieval over your documents. Vertical AI is configured by the vendor against the sector’s workflow patterns and compliance frame. Both can do useful work. They earn their keep in different ways.
When general-purpose AI is the right answer
General-purpose AI is the right answer when your work is cross-functional, your team is small, your sector regulatory load is light, and the value of integration with the productivity stack outweighs sector-specific depth.
Cross-functional teams benefit first. An eight-person professional services firm running HR, finance, operations, business development and client delivery does not need five different vertical tools. It needs one platform that works inside Outlook, Word, Excel and Teams, drafts the HR letters, summarises the financials for the board pack, and answers the team’s questions in chat. Microsoft 365 Copilot or Claude Teams at £20-£25 per seat covers that breadth at a cost no vertical tool can match.
The exploratory phase favours general-purpose too. Before you know which workflow will actually save the most time, paying £20 per user to test is cheaper than committing to a vertical contract. A practice that uses six months on Copilot to discover that proposal drafting is the highest-value use case is then better placed to evaluate whether a vertical tool would extend that win.
Light or infrequent workflows favour general-purpose. A general contractor managing five projects a year does not get much benefit from a vertical preconstruction tool optimised for a fifty-projects-a-year operation. The same holds for a generalist accountancy practice or a broad legal practice covering several areas. When the work is varied, the sector tool’s pre-tuning advantage shrinks.
Adoption friction matters here too. A copilot built into the software your team already lives in gets used. A standalone vertical platform requires people to log in somewhere new and remember to come back to it. Habit beats theoretical fit.
When vertical AI is the right answer
Vertical AI earns its premium in sectors with deep, repetitive, rule-heavy workflows and active regulatory oversight, and in firms where workflow density justifies the investment.
Legal practice with serious volume is the canonical case. A mid-market firm reviewing fifty contracts a month with Spellbook gets contract markup, citation checks and redlining inside Word, with audit trails designed against SRA expectations. A general-purpose model handed the same task is doing the work without those guardrails, and the firm has to build the governance documentation itself.
Accountancy practices with high transaction volumes are similar. Sage Copilot, embedded in Sage Intacct, automates month-end close, flags variances, proposes journal entries and tracks the close in real time. A practice processing several hundred invoices a month against a clean ledger sees close cycles compress. The ICAEW expects AI use in accountancy to carry transparency, oversight and audit trails. The vertical tool comes with those baked in.
Construction firms with active preconstruction pipelines benefit from tools like Buildr that integrate with Procore, Autodesk Construction Cloud and the BIM, scheduling and financial platforms a contractor already runs. General-purpose AI cannot read Procore APIs natively or apply construction-specific logic about labour and material sequencing. The RICS published a global AI standard in March 2026 covering governance and professional judgement in surveying. Vertical tools map to that standard; generic AI does not.
Recruitment firms running an ATS at any scale fall in the same pattern. Bullhorn Amplify and hireEZ integrate natively with the consultancy’s tracking system and keep candidate context inside the platform. The CIPD expects fairness, transparency and human review on employment decisions. Vertical tools structure that review; a generic chat interface does not.
Across all four sectors, the deeper case for vertical is professional indemnity. PI insurers in 2026 ask which tools are in use, what they are used for, what governance is in place and what vendor liability provisions cover the firm. Vertical tools come with answers designed for the sector. General-purpose deployment requires you to build the answers yourself.
What it costs to get wrong
Three failure modes show up regularly.
Buying vertical when general-purpose would have done is the first. A practice with light workflow density and broad work pays for a vertical tool whose deeper features go unused, while the team continues working in productivity tools the vertical platform does not integrate with. The result is two systems running, one of them shelfware, and a contract harder to walk away from than expected.
Buying general-purpose when vertical was needed is the opposite. A firm at the volume and complexity threshold where a vertical tool would have integrated cleanly with the case-management or ledger or ATS workflow keeps trying to make general-purpose AI work, builds complex prompts and retrieval pipelines, and ends up with a fragile in-house construction that costs more in engineering time than the vertical subscription would have. The signal is usually the team saying the AI “almost works but keeps slipping” on the tasks that matter.
Vendor viability is the third trap and it is specific to vertical. A vertical tool from a smaller specialist vendor carries platform risk that ChatGPT, Claude, Microsoft and Google do not. The vendor running out of runway takes years of workflow data with it. Vet vertical vendors on funding, customer concentration and contract termination terms before signing.
Template debt is the quieter cost. A vertical tool whose templates do not quite match your firm’s house style means hours of customisation, and vendor updates can overwrite those customisations. Confirm in the trial that the template language is close enough to yours.
What to ask before you decide
Five questions, in order.
One: how much of your team’s daily work is sector-specific versus cross-functional? If the bulk is cross-functional, lead with general-purpose. If a meaningful slice is sector-specific and high-volume, vertical is on the table for that slice.
Two: which platforms does your sector workflow already live in (case-management, ledger, BIM, ATS), and how does the vertical tool integrate with them? Native integration is the value.
Three: what does your sector regulator and PI insurer expect to see when they ask about AI use? Vertical tools have usually mapped to those expectations; general-purpose tools have usually not, and the work falls to you.
Four: what is the vendor’s viability story? Funding, customer concentration, contract termination and data-export terms. The smaller the vertical vendor, the more this matters.
Five: what does the hybrid stack cost, all in? General-purpose at the company level plus vertical for one practice area is the 2026 norm for firms with serious sector exposure. Get the combined number before signing either contract.
The honest answer for the typical UK SME in 2026 is start with general-purpose for breadth, identify the one or two practice areas where workflow density and regulatory weight justify the upgrade, and add vertical there. The right architecture is rarely all of one. It is general-purpose underneath, with vertical placed where the maths and compliance picture call for it.
