2024 AI Year in Review: The Year Professional Services Got Serious

2024 is closing out, and if I had to describe the year in one sentence: professional services firms stopped asking "should we use AI?" and started asking "how do we deploy AI at scale?" That's a meaningful shift.

Let me walk through what mattered this year, what surprised me, and what it signals for 2025.

The Big Wins This Year

Document automation became real. Contract generation, legal brief assembly, regulatory document production—these went from "interesting pilot" to "production workflow" for dozens of firms. The ROI is clear, the error rates are acceptable, and firms are scaling these systems from hundreds to thousands of documents processed per month.

Research assistance matured. A year ago, using AI for legal research or competitive analysis was risky. Now it's standard practice at firms I work with. People have figured out the right prompts, the right guardrails, and the right human review processes. The system works.

Intake and triage moved faster. Using AI to screen client intake forms, route them to the right practice groups, and pre-flag red flags—this is saving firms 10-20 hours per week of manual work. It's boring, it's not latest, and it's incredibly valuable.

The cost of AI dropped significantly. Model pricing is down 40-50% from a year ago. Smaller models (Haiku, Sonnet, Mistral) are now "good enough" for a much wider range of tasks. This expansion of the "good enough" zone is massively important for ROI.

Open source became viable. Llama 2 proved that you didn't need OpenAI or Google to have a capable model. This reduced vendor lock-in anxiety and gave firms a genuine alternative if they wanted to run models on their own infrastructure.

The Surprises

Governance was easier than expected. I thought compliance would be the big blocker in 2024. Instead, most firms figured out: "AI for routine tasks, human review for high-stakes work" is a reasonable framework. Not perfect, but workable. The governance crisis didn't materialize.

Adoption followed the early majority pattern. In 2023, adoption was concentrated among innovators and early adopters. In 2024, it spread to the early majority. Smaller firms, firms in less prestigious markets, firms with less technical expertise—they're all deploying AI now. The bell curve is real.

The hallucination problem got less scary. LLMs still hallucinate, but firms have learned how to work around it. Prompt engineering. RAG (retrieval-augmented generation). Human verification loops. None of these are new techniques, but in 2024, they became standard practice, not experimental.

Nobody talked about ethics anymore. AI ethics was a big topic in 2023. In 2024, it kind of disappeared. Not because the ethical issues are solved, but because the conversation moved from "is AI ethical?" to "how do we use AI responsibly?" More pragmatic, less philosophical.

The Disappointments

AI didn't transform client service. I expected to see major changes in how firms interact with clients—AI-powered client portals, AI-assisted client communication, etc. Mostly didn't happen. Firms were too focused on internal efficiency.

The liability framework stayed murky. Two years ago, I thought we'd have clearer legal standards by now. Nope. It's still ambiguous who's responsible when an AI-generated document has an error. This ambiguity is slowing some types of deployment.

Talent market didn't clarify. Firms need people who understand both AI AND professional services. These people are still rare and expensive. Hiring remains a bottleneck.

Small models didn't democratize as much as expected. Despite Llama and others, most firms still use OpenAI or Anthropic's APIs. Running models locally or on your own cloud infrastructure is still complicated. The democratization story is slower than I anticipated.

What This Means for 2025

2024 was the year of scaling. Firms that piloted in 2023 scaled in 2024. Now the question shifts: what's next?

I think 2025 will be the year of specialization. Instead of "we have AI," it'll be "we have AI for contract generation, AI for research, AI for intake." More targeted. More integrated into actual workflows. Less "AI as a thing we do" and more "AI as how we work."

The firms that win will be the ones that went past "deploy ChatGPT" to "build real systems." Not integrated plugins, not clever prompts—actual systems that fit into the fabric of how work gets done.

My Prediction for 2025

2025 will be the year that firms that haven't deployed AI yet start falling noticeably behind. The advantage of being early will become more visible. By mid-2025, it won't be "nice to have" anymore. It'll be competitive necessity.

For firms that are ahead: the next challenge is not implementation, it's sustainability. How do you keep people engaged with AI? How do you handle the next generation of models? How do you train 500 associates to use tools that didn't exist 2 years ago?

That's the work of 2025. Deeper. Harder. More real than 2024.