Accrual launches Arc AI agent platform for accountants

Accrual has launched Arc, an AI agent platform for accounting firms designed to handle multi-step work across existing systems.

Arc is aimed at assignments outside the standard workflows built into accounting software, such as investigating discrepancies, preparing material for client meetings, responding to tax notices, and answering questions that require information from several engagements.

Practitioners begin by giving the system an objective, Accrual said. The software then proposes a plan, seeks approval, and returns completed work for review, including outputs such as Excel workbooks, client reports, analyses, and workpapers.

The launch includes 20 prebuilt agents for common accounting firm workflows. Firms can also create their own by teaching the platform internal processes, then saving and sharing them across teams or scheduling them for recurring tasks.

The workflow gap

Accrual is targeting a category of accounting work that often falls between formal software processes for tax preparation, audit engagements, and recurring client accounting assignments. In these cases, practitioners may need to search documents, review client communications, and pull data from several systems before producing a response or analysis.

Accrual describes this as work with no dedicated screen in conventional accounting software. Arc is designed to gather information it is authorised to access, ask follow-up questions when needed, and carry out approved tasks across connected systems.

Jobs can run for minutes or hours, and several assignments can be underway at the same time. Users then review the results alongside a record of the sources, tools, and actions used during the process.

"The breakthrough isn't simply giving accountants access to a more capable AI model. It's giving that model the context, tools, controls and accounting-specific understanding required to do real work," said Siddarth Chandrasekaran, Co-founder and Chief Technology Officer at Accrual.

Firm processes

One example Accrual gave is handling a tax notice. A firm can train the platform to review the notice against client records and internal templates, identify missing information, ask questions, and prepare supporting work once a practitioner has approved the plan.

That process can then be saved as a reusable agent for other staff to use, allowing later work to start from an established internal method rather than from scratch. Teams can also schedule those agents to carry out recurring assignments.

Arc extends Accrual's wider platform across tax, audit, client accounting services, and advisory work, and is intended to work alongside the company's existing workflow products when an assignment spans several systems.

The system includes permissions, recorded actions, and human review, while final judgment and approval remain with the practitioner.

External backers and early users described the launch as a way to link existing software and operational processes rather than replace them. The product is available to current Accrual customers.

"Accrual can really become the connective layer across your work ecosystem with the ability to create agents and connect to all the applications you already use," said Alex Gyorffy, Director of Product and Innovation at Armanino.

Sector interest

The launch comes as accounting firms test where AI tools can be used beyond document drafting and search functions. The focus is shifting toward systems that can complete bounded tasks, move across multiple applications, and leave an auditable trail for review.

That need is particularly relevant in accounting, where firms must trace conclusions back to source material and maintain records of how work was prepared. Wider use of AI in the profession depends on systems that fit those review and supervision requirements.

"Accounting is well suited to this next phase of AI because the profession already knows how to get reliable work from people who need supervision," said Marc Bhargava, Managing Director at General Catalyst. "Scope the work. Tie conclusions to evidence. Preserve the record. Review what was done. Keep judgment with the professional."