Most legal operations teams evaluate AI tools before they define the business outcomes they seek, and that sequence is why governance, data quality, and staffing questions stay unresolved. Before the next vendor demo, legal ops leaders should answer three questions: what outcome the function owns, what the tool’s value would look like as a number a CFO would accept, and which judgment calls still need an accountable person. That was one of the clearest threads at Running Legal Like a Business 2026. It also points to a practical way forward: establish where the function stands today, tie AI investments to outcomes the business can measure, and design processes around what AI now makes possible.
Key takeaways
- Legal ops AI programs stall when teams select tools before defining the necessary and/or desired outcomes.
- AI governance, data quality, and organizational readiness all depend on that outcome, with each one changing depending on the goals.
- Budget pressure raises the stakes. Harbor’s 2025 Law Department Survey, conducted with CLOC across 135 corporate law departments, found that the share of departments expecting an increase in outside counsel spend fell from 58% to 37% in one year.
- With roughly 2,000 legal AI startups built on the same small group of foundation models, the tool itself rarely separates a successful legal AI program from a stalled one.
- An objective-first program follows a structured sequence: baseline the function’s AI maturity, prioritize use cases by value and risk, build the business case from the department’s own spend data, and design human supervision into pilots from day one.
- Whether AI means more or fewer legal ops jobs gets settled one function at a time while mapping which judgment calls need an accountable person.
What a weekend-built AI tool reveals about legal ops strategy
A capable legal ops professional who builds an AI tool without a defined objective will optimize for what they can see from their desk and miss what the organization needs. Consider a common scenario.
Somewhere in a corporate legal department, someone on the legal ops team is finishing an AI tool they built over the weekend. It’s genuinely good. It catches billing guideline breaches that used to slip through invoice review, and it flags clauses that have drifted off the negotiation playbook faster than a first-pass reviewer could. When you ask it to make a judgment call, it holds up.
Ask it for anything else and it gets thin. There’s no audit trail showing why it flagged what it flagged. There’s no version control, so nobody can say which iteration reviewed last quarter’s invoices. And nobody has thought about what happens when the person who built it moves into a different role or leaves the company.
It would be easy to read this as a warning about shadow IT, or as proof that legal ops teams should build their own tools. Neither is quite the point. This person built what they built because they inadvertently prioritized their own objective. Nobody had defined the outcome the function was accountable for, so they chose one they could see from their desk and optimized for it. The tool is strong on what they cared about and silent on other aspects the organization would care about.
A common debate at Running Legal Like a Business 2026
The AI concerns that dominated Running Legal Like a Business (RLLB) 2026 are symptoms of the same missing objective. At the conference the conversation kept circling three issues: hallucinations, the billable hour, and whether AI means more legal ops jobs or fewer.
Those concerns are legitimate. The root cause underneath them is that most legal ops functions haven’t defined the objective and the value they’re accountable for before they start evaluating tools. That gap explains why the workforce question still doesn’t have a good answer. You can’t answer a staffing question until you’ve answered a strategy question.
Why AI governance, data quality, and readiness depend on a defined objective
AI governance, data quality, and organizational readiness all sit downstream of one decision most legal departments haven’t made: which outcome the function is accountable for. The usual objections to moving faster on AI deserve a real hearing. Governance is hard. Data quality in most legal departments is uneven at best. Many organizations are not ready, culturally or operationally, to trust a model with work that used to require a lawyer’s sign-off.
Governance has to govern something specific: which outputs need review, what error rate is acceptable, and who is accountable when the tool gets it wrong. Those answers change completely depending on whether the goal is reducing outside counsel spend, shortening contract cycle times, or improving regulatory response.
Data quality works the same way. Clean timekeeper and matter data are crucial for spend analytics and hardly at all for clause extraction. Without a defined outcome, “fix the data” becomes an open-ended project that never quite finishes.
Readiness raises the obvious follow-up question: ready for what?
How budgets push legal departments to adopt AI before defining its purpose
Flat budgets and rising demand push legal departments to adopt AI tools before they make clear decisions on what the desired benefits are. Harbor’s 2025 Law Department Survey, conducted with CLOC (the Corporate Legal Operations Consortium) across 135 corporate law departments, shows the squeeze.
| Harbor 2025 Law Department Survey finding | Result |
| Departments expecting outside counsel spend to increase | 37% (down from 58% the prior year) |
| Departments expecting inside legal spend to increase | 47% (down from 65% the prior year) |
| Fastest demand growth: regulatory compliance | 63% |
| Fastest demand growth: cybersecurity | 58% |
| Departments with dedicated AI oversight or resources | 85% |
Under those conditions, any tool that promises greater capacity looks like a solution. That’s exactly the moment teams reach for technology, but they have yet to decide on the real impact they need the tool to deliver. The 85% figure for dedicated AI oversight is real progress on structure. Whether that oversight is attached to a clear objective is a different question, and a more important one.
How AI changes what legal operations teams can do
AI makes existing legal ops processes faster, and it also opens up objectives that weren’t practical before. Two examples show the difference:
- Reviewing every line of every outside counsel invoice, rather than sampling
- Checking an entire contract portfolio against a new regulatory obligation in days, rather than waiting for renewals to surface the issue
Neither is a faster version of the current process. Each is a different goal.
That distinction matters because knowing your workflow and knowing your objective are distinct. Workflow tells you how a process runs today. The objective tells you what outcome that process is supposed to produce, and whether a better outcome is now within reach. A team that treats mapping its current workflow as the prerequisite for choosing a tool will usually end up automating the process it already has, which caps the upside at incremental efficiency. AI allows a team to build an entirely new process around an objective it couldn’t pursue before.
Legal AI vendor selection alone won’t rescue a program
Clarity about the objective, more than the choice of vendor, separates successful legal AI programs from stalled ones. Most vendor conversations are built around workflow and skip the objective. A demo that maps neatly onto your current intake process can feel like a strong fit while leaving the more valuable question untouched.
The shakeout underway in the vendor market sharpens the point. There are roughly 2,000 legal AI startups, most of them built on the same small group of underlying foundation models. When the engine underneath is largely the same, the tooling on top is not going to be the differentiator.
Three questions legal ops leaders should answer before the next demo
A legal ops leader should walk into the next AI evaluation able to answer three questions. None of which require a long strategy exercise.
1. What business outcome is the legal ops function accountable for?
The answer is the result the business expects from the function, such as controlled legal spend, faster deal velocity, or reduced regulatory exposure. A list of the processes the function runs does not answer this question.
2. What would the tool’s value look like as a number a CFO would accept?
Value needs a baseline and a target. If the answer is “efficiency” with no baseline and no target, the evaluation hasn’t started yet.
3. Which judgment calls still need an accountable human?
Some decisions carry enough risk or require enough context that a person needs to own them, regardless of the tool’s capabilities. Others can be delegated entirely if someone is monitoring the results.
Implementing an objective-first legal AI program
Answering those questions gets a legal department to the starting line. A few deliberate steps then turn the answers into a scalable program.
- A baseline of where the function stands
Before prioritizing anything, a department needs a clear picture of its current workflows, data, people, and governance controls. That tells it which objectives are realistic now and which depend on foundational work first. - A roadmap ranked by value and risk
Use cases get prioritized by the value they would create, the risk they carry, and how ready the organization is to support them. That sequencing keeps the program from becoming a set of disconnected pilots with no measurable result. - A business case built on the department’s own spend data
Efficiency projections built on generic benchmarks rarely survive a finance review. A business case grounded in the department’s historical spend, matter volumes, and cycle times gives leadership a baseline it trusts and a target it can track. - Pilots designed for supervision from day one
This is where the weekend builder’s tool would have looked different. A governed pilot defines up front which outputs need human review, what error rate is acceptable, who owns the tool, and how each version is logged, and in doing so shows which judgment calls a person needs to own. - Processes redesigned around the new objective
If the objective is full review of every outside counsel invoice, the process around it has to be designed fresh, from who acts on the flags to how findings shape billing guidelines. Running the old sampling process faster misses the opportunity.
Will AI mean more or fewer legal ops jobs?
The workforce question gets settled one legal function at a time by leaders who have mapped which judgment calls need a person behind them and staffed accordingly. That third question is what turns an open industry debate into something a leader can actually answer. Legal ops professionals who can do that mapping, and who can tie a tool to a number the business cares about, are well positioned as AI matures.
The weekend builder had the skills to answer that question but they didn’t have an objective to answer it against. Give that same person a defined outcome, a baseline to measure against, and a governance model built for the tool they’re creating, and a weekend project becomes reliable for the whole department.
