On-Device AI, Billable Hours.
Apple's iOS 27 AI push means client data stays on-device. Here's what that changes for accounting firms running cloud-based workflows.
Apple's On-Device AI Push Means Client Financial Data May Never Hit Your Cloud
Apple's iOS 27 is building serious AI inference directly into the device — processing happening locally, not in a data center. For accounting firms, that's not a consumer story. It means your clients will increasingly be summarizing, categorizing, and asking questions about their own financial data on their phone before it ever reaches your portal or your team. The firm that doesn't have a response to 'my iPhone already told me my margins dropped' is going to spend a lot of time playing catch-up in client meetings.
If your clients' phones start answering their financial questions before you do, your value proposition just shifted — whether you planned for it or not.The Accountants Surviving AI Are Selling Outcomes, Not Hours
Rolls-Royce doesn't profit from selling jet engines — it makes money servicing them over decades. That model is exactly where accounting needs to go. The firms bleeding revenue right now are the ones still billing by the task; the ones growing are packaging recurring advisory retainers around the insights that automation surfaces. Pick one workflow your team automates in the next 90 days and immediately restructure it as a monthly advisory touchpoint instead of a line-item invoice.
Stop billing for the task automation is killing. Start charging for the interpretation it can't do.Isolated AI Workspaces Are Solving the Messy Multi-Client Data Problem
A new Python package called anvil-workspaces just hit PyPI, designed to create isolated multi-repository environments for engineering tasks. The business problem it's quietly solving is one accounting firms know well: running AI pipelines across multiple clients without cross-contamination of data, prompts, or outputs. Right now most firms cobbling together AI workflows are one misconfigured script away from surfacing Client A's data in Client B's report. Workspace isolation isn't a nice-to-have — it's what makes agentic accounting automation auditable.
Before you scale any AI pipeline across clients, ask one question: what happens if the isolation breaks? If you don't have an answer, you're not ready to scale.Your AP Automation Vendor Is About to Get Eaten by Your ERP
The standalone AP automation market — the Stamplis and Tipalti-adjacents — is on borrowed time. ERPs are shipping native agentic AP workflows fast enough that a separate vendor integration is starting to look like a liability, not a feature. When your ERP can ingest an invoice, match it to a PO, flag the variance, route the approval, and post the journal entry without leaving the system, the 'we integrate with everything' pitch from a point solution stops being compelling. One mid-market firm I know quietly dropped their AP tool last quarter after their ERP vendor shipped a native agent that handled 80% of the same volume.
If your AP automation pitch still leads with 'we integrate with your ERP,' that's not a feature — that's a warning sign.