Gusto’s new AI payroll tools could shrink the Friday email hunt

Gusto says its redesigned Gusto Pro is in early access for eligible accounting partners, with general availability planned later this year. The announced features include AI-assisted payroll preparation, Gmail and Outlook connectors for drafting hours, reimbursements and deductions, automatic recurring reports, and Pro Gus, an assistant with visibility across a firm’s clients.

A separate part of the announcement is already more widely available: Gusto says its multi-client Model Context Protocol connection is available to all accounting firms. In plain language, an accountant can connect Gusto to a compatible AI assistant once and choose which clients to include, rather than repeatedly switching individual client connections. The release describes questions across those clients, preparation of payroll changes and recurring background routines.

Gusto says every change comes back to the accountant for review and approval before it updates Gusto. Its current product page also says making changes or running payroll remains a human decision. Do not read the wider availability of the multi-client connection as general availability of every redesigned Gusto Pro feature.

For a small firm, the attraction is being able to spend less of the day finding and retyping information. Pulling the same reports for each client is an especially ordinary job to remove. An assistant that collects them in one place could help without changing how the firm decides whether payroll is ready.

The same is true of drafting from email. Finding a reimbursement message and putting its details beside the relevant payroll draft is useful preparation. Deciding whether that reimbursement belongs in this run still needs the firm’s rules and the client’s instructions.

These are product capabilities described by Gusto, not independent evidence of time saved. The announcement does not establish how accurately email extraction works on a busy firm’s real messages or how much review it removes. A firm should ask for those answers from its own trial before promising a calmer payday.

A list of proposed changes can look complete even when an expected message never arrived. That is why I would ask for a client-by-client preparation view before expanding use: the company, pay period, source of each proposed change, and outstanding inputs the firm knows it needs. This is a buying requirement, not a feature I can confirm from the announcement.

Consider an illustrative case: a client sends a revised hours file after an earlier email. A helpful preparation tool would make the relationship between the two visible so the accountant can establish which one applies. Simply putting both into a tidy summary leaves the hardest bit untouched.

The US Department of Labor’s existing FLSA recordkeeping guidance says covered employers must keep accurate records for non-exempt workers, including daily and weekly hours, wage rates, additions or deductions and the pay period. It allows different timekeeping methods as long as they are complete and accurate. That guidance is not an assessment of Gusto’s AI. It does explain why a polished draft cannot substitute for knowing which records support the pay.

For the first trial, use one client and a completed pay period, within the firm’s approved data-handling rules. Compare what the assistant prepares with the records and final payroll the firm already accepted. Do not submit another pay run as part of the test.

Look beyond matching the total. Can the reviewer identify the source of a changed figure without repeating the email search? Does a superseded message remain distinguishable from the final instruction? Can they identify an expected input that is missing? If the available tools cannot make those checks easy, narrow the trial to report gathering rather than relying on email extraction.

Then try a live preparation cycle with the normal accountant still responsible for review and approval. Count the time spent finding information and checking it, including any return to the inbox. You are testing whether preparation became easier, not whether the assistant can produce a convincing summary.

A firm does not need to adopt every new AI feature at once. If gathering recurring reports works reliably and saves a repeated chore, keep that benefit while testing the more sensitive preparation separately.

The person approving payroll should arrive at the decision with less searching left to do. If they still have to reopen every email to work out which version counts, the trial has shown you exactly where the tool needs to improve.