Making predictions about AI is a losing game

Making predictions about AI is a losing game. Every prediction from 2023 was either wildly optimistic or embarrassingly wrong. AI didn't replace lawyers. It also didn't stay a novelty. The reality landed somewhere nobody predicted -- AI became infrastructure. Invisible, essential, and boring in the best possible way.

So here's our 12-month outlook for legal AI through early 2027. Not hype, not fear -- just pattern recognition based on what's shipping now, what's in beta, and what the economics demand. Some of this will be wrong. All of it is worth thinking about.

Agentic AI: The Biggest Shift Coming

In 2026, you tell AI what to do and it does one thing at a time. By mid-2027, AI agents will execute multi-step legal workflows autonomously. Not 'generate a draft' but 'research this issue across three jurisdictions, draft a memo analyzing the results, check all citations against Westlaw, format it according to firm standards, and email it to the supervising partner for review.' The technology for this exists today in early forms -- Claude's computer use, OpenAI's Operator, and legal-specific agentic tools from startups like EvenUp and Spellbook. The bottleneck isn't capability. It's trust. Firms won't let AI agents file documents or send client communications without human approval loops. But the internal workflows -- research, drafting, document organization -- will increasingly run on autopilot with attorney oversight.

AI-Powered Court Filing: Closer Than You Think

Several state courts are exploring or piloting AI-assisted filing systems that validate documents before acceptance, check formatting compliance, and flag potential issues. By 2027, expect at least a handful of courts to offer AI-powered filing validation that catches errors before they become problems. This benefits lawyers too -- no more rejected filings for formatting issues, missed deadlines discovered at 4:59 PM, or service failures caught weeks later. The more radical prediction: at least one court will begin accepting AI-generated standard filings (certificates of service, proposed orders, routine motions) through automated portals. The paperwork of litigation is ripe for automation, and courts are as resource-constrained as the lawyers appearing before them.

Client-Facing AI: The End of Information Asymmetry

By 2027, your clients will arrive at consultations already informed by AI. They'll have asked ChatGPT about their legal issue, gotten a rough analysis, and formed expectations before they ever call your office. This isn't a threat -- it's an opportunity. The lawyers who adapt will use client AI knowledge as a starting point: 'I see you've researched this. Let me explain what the AI got right, what it missed, and why your specific situation is different.' The firms that fight it -- dismissing AI research, refusing to engage with client-generated analysis -- will lose clients to firms that meet them where they are. Client education is shifting from 'let me explain the law' to 'let me explain what AI couldn't tell you.'

Pricing and Billing: AI Forces the Conversation

The billable hour model is under pressure from AI efficiency, and 2027 is when it starts cracking. If AI cuts a 10-hour research memo to 2 hours of work, billing 10 hours is indefensible and billing 2 hours leaves money on the table. The answer is value-based pricing, and AI makes it possible because the cost of delivery becomes predictable. Expect more firms to offer flat-fee arrangements powered by AI-assisted efficiency, alternative fee structures that price outcomes instead of hours, and transparent AI efficiency disclosures to corporate clients who are already asking 'are you using AI to reduce our costs?' The firms that proactively restructure pricing around AI efficiency will win the clients that care about value. The firms that hide AI efficiency to protect hourly billing will get caught.

Regulation: The Rules Are Coming

2027 will bring the first comprehensive AI regulation affecting legal practice. The EU AI Act's provisions covering 'high-risk AI systems' will apply to legal AI tools used in court proceedings. US state bars will continue issuing opinions, but expect at least one state to adopt formal rule amendments (not just opinions) addressing AI in legal practice. The ABA will likely propose Model Rule amendments specifically addressing AI -- moving beyond guidance into binding ethical rules. The firms that have been treating AI ethics as optional will scramble. The firms that built compliance infrastructure early will have a 12-month head start on competitors. This is the regulatory cycle every technology goes through: experimentation, guidance, regulation. We're at the end of the guidance phase.

The Bottom Line: The next 12 months will bring agentic AI workflows, increased regulation, pricing pressure, and AI-informed clients. None of this should surprise firms that have been paying attention. The window for 'wait and see' is closing. By 2027, AI adoption won't be a competitive advantage -- it'll be a competitive requirement.