AI Thrives on Order, and So Does Your Business

As insurance leaders experiment with artificial intelligence, many are approaching it the same way they do every new tool: layering it onto existing workflows such as email and contract review as an optional resource.

That instinct to ease in to using AI is understandable. It’s also limiting.

AI can be applied to something far deeper than what McKinsey calls “fragmented use cases.” It offers an opportunity for the industry to rethink how work is designed.

When we evaluate what automation and AI can handle, it forces a broader question: why are we doing this job this way in the first place? This process made sense a few years ago, but does it still add value?

For several months, I’ve been working alongside new GenuineShift “team members” Claudine (Anthropic’s Claude), Jared (Google’s Gemini), and Leo (ChatGPT via OpenAI). I’ve used them as data analysts to compare outcomes from each Academy cohort, as advisors on our long-term business strategy, and to respond to my favorite prompt when I’m writing, “Make this shorter.”

Each has distinct strengths and approaches the work differently. None get the job right the first time or fix anything without clear direction.

When the output misses the mark or requires repeated clarification, it’s not just friction. It’s feedback that our thinking or process isn’t clear enough to produce consistent results. That’s frustrating enough at the individual level, but across a team or division, that lack of clarity multiplies. AI only amplifies inconsistencies that are holding back your team.

Understanding Use Cases

In one case, I wanted to update my CRM to show each contact who is attending The Council’s Employee Benefits Leadership Forum (EBLF) this year. It should have been a simple task for the AI—find the contact and add the “EBLF ’26” tag; instead, the process became complex. The tool evaluated existing data, attempted to optimize the logic by predicting other attendees, and produced the wrong outcome when it removed prior tags entirely and hallucinated 10 attendees from mysterious new companies.

The issue wasn’t capability, it was clarity. When I didn’t give specific instructions, it hallucinated, filled in gaps, and time and energy were lost correcting the result.

In GenuineShift training, we talk about “clarify and verify.” Direction that is too vague or overly rigid can get you into trouble with both AI tools and people, creating unnecessary complexity, incorrect execution, or both. Don’t assume others understand what you expect. Be clear and specific.

Understanding the use cases for AI, automation, and human judgment is also important. In one instance, I unleashed Claude Cowork to individually analyze data in our system when the less-powerful chat tool could have more quickly and efficiently compared two spreadsheets. The result was unnecessary complexity, rework, and AI token costs.

Running out of tokens is a clear signal of a chatbot job done wrong. When I exceed limits in one Cowork session, I am prompted to approve a surcharge on my Visa. There are no such prompts when you reach the limits of your team’s capacity. But, given that one of the biggest issues facing agencies today is lack of enough people to do the work and scale with their growth, we must do a better job of resource allocation. The other option is missed deadlines, reactive work, and client frustration.

Every day in insurance operations, leaders ask whether someone can completely handle a job. That’s the wrong question. Most work doesn’t require a single owner. The better question is how to structure the job to be done.

It requires the right combination of internal team members, external partners, automation tools, and now AI. The right structure avoids inefficiency and creates capacity.

Go With The Workflow

In many teams, the steps required to complete a task live in someone’s head—the follow-ups, the workarounds, the last-minute checks to make sure everything is correct. That invisible work becomes the system until it breaks under pressure.

You can see this clearly in renewal workflows. In many agencies, each producer or account executive approaches renewals differently. The result is inconsistent tracking of open items and variation in materials sent to clients.

That forces the account manager to deal with differences in timelines, expectations, and communication styles for each renewal. In some cases, they are doing that with multiple business partners and clients, involving different producers or account executives.

What looks like flexibility at the top creates complexity underneath.

These workflows are often treated as optional, shaped by preference, experience, or individual working style. The work product (claims analytics, marketing results, or employee communication materials) may still be strong. A skilled account manager can pull it together, even if it requires a bit of duct tape.

We don’t have a surplus of experienced account managers. When each renewal requires them to convert an individual approach into a consistent process, we’re not just creating variability—we’re introducing cost.

When workflows are optional, execution becomes unpredictable and unprofitable.

One of the most useful concepts I’ve taken from working with AI is designing for repeatable outputs. It’s about structuring a process to produce consistent, reliable results every time.

In insurance terms, that means moving from “how this producer likes to run a renewal” to a defined process that can be executed across clients, teams, and tools. Once that process is clear, it can be supported by automation, enhanced by AI, and scaled without requiring reinvention each time.

One Step At A Time

One of the most practical lessons learned from working with AI is the importance of small-batch execution. Instead of running an entire workflow, test a small portion—10 rows of a spreadsheet, a subset of records, a partial output. Validate the result, adjust direction when necessary, and then expand the job.

This is a well-known but often ignored idea. Without checkpoints, team members get well down a path researching or developing something only to find out they committed to a certain approach too early and must spend more time correcting than executing.

Questioning my own business’s approach to certain work led to a redesign of core processes around what automation and AI tools could handle effectively.

This shift has practical implications. We were hitting a 25-hour weekly capacity cap with one external partner. Redesigning our approach to updating CRM records and other tasks created additional time for higher-impact tasks including improving client communication and engagement.

The takeaway is straightforward: AI does not fix broken systems, it exposes them.

Whether you are working with Claudine, Jared, Leo, or your own team, your most valuable resources remain the same: time, attention, and capacity. How you design work around those resources ultimately determines what you can deliver for your clients and your business.