The Fastest AI Still Moves at the Speed of the Firm

PI firms have an incentive advantage, but not a human exemption

Your firm has an advantage that many billable-hour practices do not. Completing routine work faster does not usually reduce your revenue. Faster response, stronger conversion, and fewer stalled cases can improve both the client experience and the economics of the firm.

That alignment helps, but it does not settle the incentives inside your team. An intake specialist may be measured by calls handled, not meaningful contact. A case manager may be rewarded for closing tasks, not surfacing a difficult exception. An attorney may see personal responsiveness and judgment as central to professional identity. Your operations leader may worry that automation will expose inconsistent data or undocumented processes.

None of those concerns is irrational. They tell you something important about the environment into which you are introducing AI.

If you say, “Use AI,” while promotions, recognition, and performance reviews still reward the old activity, your team will follow the incentive. If you describe AI mainly as a way to remove headcount, people will protect the work that proves their value. If attorneys receive polished answers with no source path, they will distrust them, and they should.

The second mismatch is professional identity

A vendor sees faster summaries, automated actions, better retrieval, and broader coverage. The person expected to use the system may see uncertainty about expertise, career progression, and status. You need to understand both views.

Elizalde-Bulanti offers a useful question to ask each person: which part of your work, if removed, would make you feel less valuable? Ask it before you redesign the workflow, not after you announce the rollout.

For an intake specialist, the answer may be the ability to calm a frightened prospect and earn trust. Your system should prepare and accelerate that human interaction, not replace it. AI can structure the inquiry, identify missing facts, detect urgency, and brief the right closer. A person still creates the relationship. That is the philosophy behind our approach to human-led AI intake automation.

A case manager may find value in understanding the client and recognizing when a file does not fit the usual pattern. Let AI maintain the chase queue, draft routine follow-up, and assemble the treatment timeline. Let the case manager own the exception, verify uncertainty, and decide when a client or attorney should become involved.

For your attorneys, AI may prepare source-linked work. Advice, strategy, case acceptance, valuation, settlement authority, and consequential client conversations remain human responsibilities. Your team will adopt the system more readily when you state that boundary clearly.

Co-design produces adoption that training cannot

Elizalde-Bulanti invokes the IKEA effect: people tend to value what they helped build. Yet firms often do the opposite. A small group selects a product, consultants map the workflow, and users first see the finished system during training.

By then, you have already made the decisions that matter. Your staff can learn where to click, but they did not help decide what the system should notice, what a good answer looks like, when it should pause, or where an exception should go.

Bring the people who perform and supervise the work into the first design conversation. Your intake specialist, case manager, paralegal, attorney, and operations leader will see different failure modes in the same workflow. You need all of those views before hierarchy decides what matters.

This is why a working session is usually more useful than a broad AI lecture. Let your team see live work early, test it against familiar examples, disagree with it, and help improve it. Our workshops for PI teams are designed around that principle.

Your clients belong in discovery too. If you are automating updates, ask injured clients which moments created anxiety, which explanations were unclear, and when they needed a person. Legal service can learn from healthcare, hospitality, and other high-trust environments.

Safe experimentation is a management decision

You cannot improve a new workflow without experimentation, and some experiments will fail. In legal operations, the boundary matters. A failed test with sanitized data is learning. A missed deadline or an unreviewed communication affecting a client is not.

You can create room to learn without becoming casual about risk. Start with synthetic or sanitized matters. Keep permissions narrow and actions reversible. Preserve the source behind each output. Stop low-confidence or consequential cases for review. Record overrides, errors, and exceptions so the next version improves.

Those controls do more than protect the client. They make adoption feel safer to your team. People are more willing to test a system when you have defined what can go wrong, who owns the exception, and how a mistake will be corrected. Clear AI governance and vendor-risk controls belong inside change management. They are not paperwork to add after the real work.

ROI matters, but it is not your first test

An AI system can produce an impressive answer and create no value. The question is not simply whether the model worked. It is whether the workflow changed.

Elizalde-Bulanti's point is not that ROI does not matter. It is that asking for mature ROI before your team has learned how the new workflow should operate can kill a useful initiative too early. Early on, you are testing whether the firm can learn, adapt, and use the system responsibly. Financial return becomes a fair test once that behavior begins to stabilize.

Your early measures should show whether people are using the new path: completion rate, review time, override rate, exception volume, source-verification failures, and confidence reported by users. Those are leading indicators, not vanity metrics.

Then measure the business result. For intake, look at time to meaningful human response, contact rate, retainer delivery, and signed cases. For case development, look at outstanding records, treatment-gap visibility, aging work, cycle time, and client-update burden.

This is not permission to avoid accountability. It is a better order of operations: measure learning first, changed behavior second, and business value as the workflow becomes reliable.

A practical operating model for PI firms

Start with one workflow. A firm-wide instruction to adopt AI is too broad to manage and too vague for your team to act on.

What this looks like in serious-lead response

Suppose a web inquiry describes a significant injury. In the old workflow, it enters a shared inbox, waits for review, and is handled in arrival order. Meanwhile, the prospect keeps calling firms.

Your goal is not to “let AI conduct intake.” Your goal is to recognize a serious inquiry immediately, prepare the right person, and make fast human contact.

Let AI structure the submitted facts, identify missing information, flag urgency or shopping signals, and prepare a concise brief. Your intake specialist confirms the facts. The attorney or designated closer receives the context and contacts the prospect. Case acceptance remains with your firm.

The change-management work sits around that sequence. Who helped define a serious lead? Does your closer trust the brief? What happens when information conflicts? Do you reward the intake specialist for escalation quality or merely queue volume? Can you see whether faster response improves meaningful contact and signed cases?

The model is only one component. The transformation is the reliable human system you build around it.

The durable asset is adaptability

Models will improve. Vendors will add features. Interfaces will change. If you treat every release as a fresh technology rollout, your firm will remain dependent on outside momentum.

The more durable assets belong to your organization: a clear workflow owner, documented authority, trusted data, approved examples, evaluations, review habits, feedback loops, and a team that knows how to improve a system without abandoning it or trusting it blindly.

This is why AI readiness begins before AI. Your real advantage is not early access to a model. It is the ability to absorb changing technology while preserving judgment, reliability, and client trust.

The fastest AI available cannot outrun your organization. You do not solve that by demanding that people move at machine speed. You solve it by creating an environment in which your people and the technology can do better work together.