The following is a guest post from Konstantin Dzhengozov, CFO and co-founder of Payhawk. Opinions are his own and not those of CFO Leadership.
For the past two years, financial technology vendors have promised self-driving ledgers and autonomous closes. That has left many CFOs to assume that vendors would weld pivotal AI innovations onto their ERP suites.
However, the most practical incarnation of finance AI is emerging where day-to-day decisions are already being made: in the tools teams use to collaborate. That means Slack. Or Teams. In short, the “places” we all work.
Technological maturity is quickly diminishing the standard objection that chat isn’t safe for finance. Consider: Slack’s Workflow Builder and Microsoft’s Graph APIs now have secure hooks for file uploads, identity checks and writebacks to back-office systems.
A finance bot can now surface a purchase order in a channel, verify the approver’s role via single sign-on and record the outcome straight into the ERP, with no unsecured screenshots and no shadow IT. In another example, modern agent frameworks can read company policies, respect role hierarchies and create a clear audit trail for regulators.
Early adopters are discovering that the payoff is measured less in autonomy and more in ruthless cycle-time compression. Previously, they relied on dashboard nudges and weekly mass emails, but results were still slow.
But once they layered in Slack-based reminders, missing documents came in within hours instead of days, easing the month-end scramble and freeing accountants from constant chasing. Bots flagged validated expense claims for one-click approval and nudged account managers the moment invoices passed due (replacing overnight batch delays).
The accumulated minutes saved across these micro-decisions unlocked real strategic gains.
Inserting Agents
Gartner expects that by 2027, a third of all enterprise software will ship with embedded agents, and organizations will make up to 15 percent of their everyday business decisions autonomously. Yet, many organizations remain bogged down in heavyweight AI projects aimed at automating the entire close process.
The lower-risk, faster-return route is to insert agents at the collaboration edge, where integration hurdles are lower and the human-in-the-loop is already staring at the screen. That means CFOs will do well to start with the plumbing, not the palace. Here’s how:
- Clean. Clean the data pipes that feed your spend and approval workflows.
- Test. Expose those data pipes securely to your chat platform. Let small, deterministic agents prove their worth chasing receipts, enforcing policies and surfacing live KPIs in the daily stand-up.
- Expand. Once that lattice is in place, more sophisticated reasoning (like forecast variance explanations and real-time scenario modeling) has a trustworthy stage on which to perform.
Five years ago, the finance team’s window to the world was still the monthly Excel pack. This year, it will be the blinking cursor in a chat thread. While the ledger will remain the system of record, the chat window is fast becoming the system of action. Ignore that migration, and tomorrow’s autonomous enterprise will happen one line at a time in your competitor’s channel instead of yours.
This column first appeared in the June 13 issue of CFO Leadership’s Finance & Accounting Technology Briefing.
