Sage has made the case for agents in the most deadline-driven corner of UK accounting. In How agentic AI is changing accounting compliance, the company argues that Making Tax Digital is moving compliance “from inputting data to reviewing it,” and that agentic AI can “suggest transaction categories at source, flag possible errors before submission, support the workflow, and help give clients insight, with accountant review at each step.” The pressure is real and dated: since April 2026, sole traders and landlords with qualifying income over £50,000 file quarterly under MTD for Income Tax, and the threshold falls to £30,000 in April 2027 and £20,000 in April 2028.
The product behind the argument is the Sage MTD Agent, which segments clients by MTD readiness, drafts quarterly reports from digital records for review and HMRC filing, and chases missing documents “via email, WhatsApp or in-app notifications.” The headline figure, “up to 80% less admin,” is a vendor number with no method attached, and should be read as one. What survives it is the design principle the post repeats in every section: the accountant approves what is sent and submitted. “Control over automation: You stay responsible for what’s submitted.” That is the right line to draw for an agent that touches a tax filing.
The catalog answer is that the agent sits on top of a substantial API estate, and the estate stops before the agent does. The Sage provider page lists 13 API pages, and the ones this story leans on are there: the Sage Sales Invoices API, Sage Purchase Invoices API, Sage Bank Accounts API, Sage Ledger Accounts API and Sage Tax Rates API are the digital records a quarterly update is drafted from. The Sage agentic access profile counts 25 operations, 12 of them acting. None of the API pages describes the MTD Agent itself, client segmentation or the quarterly submission step. The Sage Accounting record, the UK product this post is written for, holds one page, the Sage Accounting v3.1 REST API.
Sage scores 40.4 on the Kin Score, developing band, carried by contract quality at 66.4 and developer ergonomics and discoverability at 57.1 each, with access clarity at 5.3 and operational transparency at 7.9. Agent Readiness is 29.8, with error semantics and delegated identity lit and dry-run mode, reversibility and idempotency unlit. Sage Accounting scores 20.7 with Agent Readiness at 2.5, human-only. For an agent whose whole promise is “flag possible errors before submission,” the unlit dry-run and reversibility dimensions are the pointed ones: Sage has told accountants they stay in control, and its public record does not yet show a developer how a submission can be previewed or undone.