Close management software has traditionally helped firms bring order to a demanding process. It standardises working papers, checklists and procedures, assigns responsibilities and gives teams a shared view of progress across a portfolio of clients. That has been valuable but the software has largely tracked the work rather than performed it. Accountants still collect the data, update the workbook, investigate movements and prepare journals. The platform co-ordinates people under time pressure but the work itself remains manual.

Agentic software changes that model. Collaboration is no longer only human to human. It also becomes human to agent with accountants directing, reviewing and approving work carried out by software agents. In practice, this does not require firms to replace the Excel workbooks around which many close processes have been built. An existing workbook can be digitised into a close template by extracting checklist items, matching each item to one or more working papers and defining the dependencies or prerequisites between them.

Close management is evolving from co-ordinating people and processes to orchestrating work performed by accountants and software agents.

‍Each checklist item can then carry a structured prompt covering three areas: the data to collect, the analysis to perform and the expected output pattern. Agents retrieve information directly from core accounting systems, complete work within the workbook and prepare it for review. Runs can be scheduled overnight so updated working papers are ready when the team starts in the morning. Changes are recorded in an audit log, explained and ranked by materiality. Where an adjustment is required, the agent proposes a journal for human approval rather than posting it autonomously.

What makes the system dependable?
‍
For CIOs and engineering teams, the most important questions are rarely about which underlying model is used. The harder engineering lies in the harness around the model: how work is structured, how access is controlled, how outputs are validated and how the close continues when part of the technology stack fails.

Source documents are a good example. Accounting data often arrives in inconsistent formats, including invoices, contracts, schedules and client-specific reports. Agentic document extraction can interpret these sources within the context of the task being performed. Firms can also define custom document types that retain the appropriate accounting treatment, reducing the need to rebuild the same instructions each month.

Accounting data often arrives in inconsistent formats, requiring systems to interpret documents in the context of the task being performed.

Resilience matters just as much. A period-end close cannot depend on a single model provider. A multi-model architecture can route work between the highest performing foundation models, with alternative models ready to take over if a provider becomes unavailable, so an outage does not disrupt time-sensitive client work. For UK firms, data residency can also be maintained through a London-based tenant.

Does the system learn from the firm?
‍
A useful system should improve as reviewers correct its work but that does not require model fine-tuning or the transfer of a firm’s intellectual property into a shared model. After the close, a review pass can capture the changes made by accountants and store them as additive context in a per-client intelligence layer. That intelligence remains accessible only to the firm and can inform how the same task is approached in future periods.

This is the direction we are pursuing at Artifact AI: not replacing the close process with an entirely new platform but adding a controlled execution layer to the tools firms already use. The next generation of close management will not simply show teams how the work is progressing. It will help complete the work while keeping accountants firmly in control.

Alexander Crump, ACA, is a former KPMG auditor and UK GTM Lead at Artifact AI, which deploys specialised AI accounting agents into leading UK and US accountancy firms.