In brief
Mode's BizBuyData program targets accounting firms with at least 10 employees and records they have rights to license.
The vendor says it removes sensitive information before onward delivery; pricing and timing depend on the written agreement.
Digits and Clark are expanding AI from answering questions to taking actions inside accounting workflows.
A new buyer for operating history
Mode is offering accounting firms money for copies of their operational records through BizBuyData, a program intended to supply data for AI research and training. The proposition is straightforward. The rights review is not.
The vendor's current website identifies accounting and finance businesses with at least 10 employees as potential participants. It lists records such as email, documents, chat, customer-management information and billing data, and says companies with several years of records they own are the strongest fit.
Mode says the parties agree on systems and date ranges before export. It then prepares a copy, removes sensitive material and removes or transforms personal identifiers before delivering the filtered data for AI research and training. Original records and systems stay with the participating company.
Do not confuse possession with permission
An accounting firm may hold a record without having unrestricted rights to license it. Engagement terms, client confidentiality obligations and the nature of the underlying information need review before a firm agrees to supply data.
De-identification is a vendor-described process, not a substitute for deciding whether the firm can provide the material in the first place. Partners should establish which records are eligible, which must be excluded and how the agreed restrictions will be documented.
The website's payout calculator is illustrative rather than a guaranteed offer. Mode says the written agreement sets the amount and payment schedule, and currently describes roughly three months from the first conversation through payment. A firm should evaluate the actual contract, not build a revenue forecast from a marketing estimate.
AI gains write access to the books
Digits has introduced Write MCP, allowing compatible AI agents to change ledger information rather than only retrieve it. The company's site confirms the update is protected by its Locked Periods feature.
The reported capabilities cover transactions, bills, invoices and chart-of-accounts records. Agents operate under the signed-in user's permissions, while actions involving locked periods require human review and approval.
That changes the evaluation question. A useful demonstration is no longer just whether an assistant can answer a financial question correctly. Firms need to see which changes it can make, how closed periods are protected and how reviewers identify an incorrect entry.
Audit software becomes more active
Clark's Material Intelligence product is designed to monitor an audit as new files and communications arrive. The company says it can identify missing information, request it from a client, track the response and resume work after the material arrives. Auditors remain responsible for professional judgments, review and sign-off.
These developments should not be treated as one undifferentiated AI upgrade. Licensing records for model training, giving an agent ledger permissions and automating audit follow-up create different obligations.
For firm leaders, the practical next step is to identify the data involved, the authority being granted and the human decision that remains. Evaluate those specifics before accepting either a revenue promise or a productivity claim.
