AI Gets Cheap. Workflow Gets Hard.

Morning. Damian here — technically his digital twin. Same voice, same opinions, less blinking. DayLift Signal. AI-curated. Five minutes.

AI is getting CHEAP. Your strategy just got harder. I went through the Friday pile... model chatter, feature bait, the usual. This is the one shift that actually changes how US tax and advisory firms should plan next year.

OpenAI, Anthropic, and Google are all pushing prices down and billing flexibility up. OpenAI already cut GPT-five point six Sol pricing, Anthropic and Google are answering, and Google Cloud is adding spend caps, project controls, and off-peak discounts for enterprise usage. The REAL story is not smarter chatbots... it is AI turning into a utility line on your P and L.

For the Solo or small tax and accounting practice, that is good news only if you stay narrow. Cheaper AI does not mean buy more tools. It means lock in one or two ugly, repeatable jobs that save owner time before busy season — notice-response first drafts, document chasers, internal summaries. For the Multi-person accounting and advisory firm, this is bigger — realization, rollout, and usage discipline. If ten, twenty, or forty people can hit premium models whenever they feel like it, your margin leak gets faster... not better. You're still treating model choice like strategy when the real edge is deciding which work should even touch AI at all. Independent financial advisor or R I A or wealth manager — useful backdrop, yes, but I am naming the skip a bit today. The sharpest move here is internal workflow economics, not S E C or FINRA-governed client communications. Smart move: run a twelve-month AI budget now, based on lower prices, and choose two or three workflows you will standardize on purpose.

Here is the lever. This one's for firm owners and team leads first. Use a three-layer AI plan.

Layer one: standardize no-regret work now. Internal email drafting, checklist creation, meeting recap drafts, non-client-data summaries inside Microsoft Copilot or an approved business tier. Layer two: ring-fence client-adjacent experiments for ninety days. Tax research support. Planning memo drafts. Review summaries. One owner. One metric. Layer three: hold the high-risk stuff. Return positions, suitability recommendations, AI marketing at scale under the S E C marketing rule. Keep client data out of consumer tools, and stay inside business systems with retention controls and logs. First step today: write your three layers on one page.

Here is my honest take... model rankings now change so fast they are becoming a distraction. The firms that win next year will not be the ones bragging about the smartest model. They will be the ones managing work better — what gets batched, what gets routed cheap, what still gets human review. AI is starting to change management, not just tasks.

The trap is trying to design one perfect, firmwide AI rollout. I see this in mid-sized firms all the time, and solo shops do a smaller version of it. One giant knowledge base. One huge assistant. One all-or-nothing decision. Meanwhile, staff keep using shadow tools with no standards, no logs, and no link to billable reality.

Better frame: treat AI like a portfolio, not a moonshot. Run a few small bets in parallel. Kill some. Double down on a few. If the project has to be perfect before it starts, it is probably NOT a real operating move.

So here is the question. Over the next twelve months, which two AI workflows in your firm are you deliberately backing — and which tempting projects are you explicitly saying no to?