The latest AI news reinforces a practical point for firm leaders: better results are less about asking a general model to do more, and more about building the right workflow around the right data. That is especially true in law and accounting, where accuracy, verification, and repeatable processes matter more than novelty.
AI investment is growing, but infrastructure is the bottleneck behind the scenes
One recent report says global investment in AI infrastructure is projected to reach $31.6 trillion through 2050, with the US expected to capture almost half of that investment. The same report says data center spending is forecast to rise over time and that chips and other ICT equipment will need regular upgrades.
For professional-services firms, the takeaway is not to chase infrastructure headlines. It is to recognize that AI is becoming more expensive and operationally serious, which makes selective deployment and clear workflow design more important than broad experimentation.
Why generic AI use is delivering limited time savings
A Deloitte survey cited in the legal press found that many employees use AI at work, but most see only small productivity gains. The common uses were searching for information, summarising information, and finessing emails.
That pattern matters for firms because those are useful support tasks, but they are not the same as a firm workflow. If AI is only helping people search or draft a note, the time savings can stay modest. The bigger opportunity is to build AI into repeatable steps such as intake, matter setup, tax research triage, document review support, and status updates.
In legal AI, the data layer may matter more than the model layer
Another recent legal AI article argues that when model quality improves across the market, differentiation shifts to the data layer. In legal work, the underlying material has to be current, structured, and reliable if the output is going to be useful.
That point applies to both law and accounting. A custom AI workflow is only as strong as the documents, matter data, engagement history, tax references, or policy content it can actually use. Firms that organize their own information well are better positioned to get useful, verifiable output from AI tools and agents.
Custom automation is where firms can turn AI into operating leverage
The most practical response for firm owners is to focus on specific workflows rather than general AI adoption. In law firms, that could mean intake and triage, document summarisation, or internal research support. In accounting firms, it could mean client intake, request tracking, or review preparation.
Agentic workflows are especially useful when work needs a sequence of steps, not just a draft response. The value comes from connecting the right inputs, checking the right sources, and handing off to staff at the right point. That is where custom AI starts to look like operating leverage instead of a software experiment.
- Start with one repeatable workflow where time is lost to searching, summarising, or handoffs.
- Treat firm data as part of the product; clean, structured information makes AI more useful and safer.
- Measure AI by completed work and reduced friction, not by tool usage alone.
Sources watched
- Global Investment in AI Infrastructure to Hit $31.6 Trillion Through 2050 (CPA Practice Advisor AI)
- AI Gains Slim for Most Staff, Is Legal Different? (Artificial Lawyer)
- When Every Legal AI Has a Good Model, What Differentiates It? (Artificial Lawyer)
