News You Can Use

Despite constant talk of an AI bubble, a collapse in 2026 seems unlikely. Most legal AI vendors are still reinvesting aggressively in compute, talent, and infrastructure rather than extracting profits, making the market less a pyramid scheme and more an ongoing fire consuming resources to meet demand.

This growth model brings risks: energy use, data centres, environmental impact, and waste. These pressures are more likely to trigger pushback than a sudden burst. In 2026, the challenge for legal AI won't be failure, it will be the cost of continued expansion.

Early adopters of legal AI will have moved beyond experimentation, with measurable cost and time savings beginning to show. These gains make it easier for clients to compare firms, while integrated AI platforms and data-driven workflows signal the rise of Legal Ecosystem 3.0.

Runway runs out for a big chunk of legal/AI startups as markets settle around category winners. Funding doesn't vanish but becomes ruthless: ARR quality, retention, and integration depth matter more than vision decks. Success becomes rarer and more technical. Expect fewer new entrants making headlines and more activity around M&A and acqui-hires as struggling startups consolidate into the emerging category leaders.

Corporate legal teams invest heavily, driven by pressure to do more without headcount growth. AI-native vendors and ALSPs start absorbing low-complexity work. Legal engineers appear inside corporate teams, not just in firms.

Expect noticeable capability gaps between AI-mature in-house teams and everyone else. The in-house/outside counsel relationship will continue to shift as clients build more of their own tooling and scope out only the work that genuinely needs a firm.

The real progress happens in embeddings, SLMs and orchestration rather than frontier models. Model orchestration, the ability to select the right model for each task and switch between them, may emerge as a key differentiator, allowing providers to offer flexible, purpose-built solutions as a unique selling point in the market.

Firms that treat "which model for which task" as a design problem rather than a procurement question will extract more value than those chasing raw frontier capability.

Knowledge agents emerge that surface prior advice, precedents, and matter reasoning. Clients are keen to build their own agents and processes, supported by a growing ecosystem of tools, while firms adopt more cautiously. Vertical agents take shape in regulated domains (e.g. legal, finance, health) offering structured, sector-specific support. Within firms, triage agents transform rough AI-generated or client-submitted drafts into actionable outputs, saving lawyers' time.

Direct-to-client agent delivery remains limited, but internal applications for workflow efficiency are expanding rapidly. The firms that get ahead on agent orchestration inside the firm will be best positioned when client-facing agents mature.

Every major platform folds agents or AI-native features into its core workflows. The standalone "AI tool" era fades. These in-platform / ambient agents often wrap APIs and automate tasks that previously would have required advanced technical knowledge or "super user" skills, making powerful functionality accessible to all users.

Providers can deliver more seamless, task-oriented support, helping users complete specific workflows efficiently while keeping AI firmly integrated into the broader platform experience. Standalone legal AI products will need to differentiate hard on workflow depth or risk being subsumed by platform-layer features.

Search, shopping and general research become AI driven by default, and legal research isn't far behind. At the same time, a counter-movement grows around "true" unfiltered results. SEO adapts.

Business development efforts will begin designing strategies specifically for AI and assistant-driven discovery, a trend likely to be most relevant in more commoditised areas of law where ease of access and visibility are key competitive factors. Firms not thinking about how their content surfaces inside AI assistants will lose visibility they currently take for granted.

Regulators are starting to intervene more directly, as seen in AI Growth Lab requests and increasing attention from the SRA and Law Society. While frameworks aren't fully in place yet, firms are under growing pressure to clarify data provenance, training sources, and risk controls.

Any high-profile failure could trigger a surge in scrutiny, making governance a key bottleneck for otherwise-viable AI deployments. Firms that invest in provable governance now buy themselves speed to deploy later.

Lightweight, automation-driven firms are emerging (e.g. norm.ai, Garfield, Soxton), targeting highly repeatable legal work. They may take share from traditional firms in commoditised segments but are unlikely to enter top-100 work requiring depth, judgment, and brand. ALSPs are beginning to offer AI-native offerings into their services, and some traditional firms may follow where margins make sense. Tech vendors, such as Harvey or Legora, could inadvertently carve this market by giving in-house teams self-serve AI tools, though they are unlikely to formally become law firms.

Regulators are keeping a close watch on this evolving space. Traditional firms should track which segments genuinely erode and which stay defensible by judgement and brand, rather than reacting to every new AI-native launch as an existential threat.

Google's infrastructure stack, distribution and data gravity give it the edge as the market matures. Its models get paired with cloud, Workplace and Android integration in a way OpenAI can't match without burning cash. OpenAI appears to be pivoting toward monetisation strategies focused on advertising, shopping, and other revenue-generating avenues, reflecting the need for sustainable income. Google, by contrast, can leverage its diverse revenue streams to maintain and expand its dominance over a long horizon. Anthropic stays relevant or gets acquired by Google.

Competition will continue, but the scale and depth of Google's ecosystem give it a clear upper hand in the market. For firms, this is a prompt to avoid strategies that bake in a single named frontier lab and instead build provider-agnostic workflows.

Firms need new ways to build judgement and muscle memory. Simulators such as transaction walkthroughs, interview role-plays, drafting under pressure become mainstream. Similar to training pilots: high-fidelity, high-stakes rehearsals without real-world risk.

This approach ensures that when lawyers face real-world challenges, their responses are precise, confident, and well-trained, bridging the gap left by AI-driven task automation. The apprenticeship question raised by agentic AI gets partly answered through deliberate practice environments rather than on-the-job exposure that AI has removed.