Skip to content

Share

The AI Playbook for Mid-Market Leaders

An AI strategy starts with the services customers will pay for as technology changes what they can do themselves. That assessment guides the choice of workflows to redesign and the pilots worth funding.

A production worker uses a tablet to review operations beside a food manufacturing line.

Where AI changes the economics of the business

Consider an insurance broker whose small business clients can now compare policies themselves. Automating renewal emails may save time. It leaves a larger question unanswered. Which clients still value the broker’s advice enough to renew?

Zenith’s Integrated AI Framework brings those questions together. The strategy assessment tests where industry profits might move and which advantages the business can defend. The value creation plan translates that assessment into changes to individual workflows. Management sets the controls for both.

Exhibit 5 The Integrated AI Framework

Which parts of a workflow can AI handle?

A distributor’s service team can check an order record and a warehouse update before telling a customer when a late delivery will arrive. A request for compensation calls for a different level of judgement. Assess the tasks separately before deciding how much of the workflow to hand over.

The assessment below rates volume, rules, ease of checking, stakes and judgement. Use the result to identify where a person should retain control, where AI can assist and where an automated step can be checked reliably.

  1. Anchor

    Typically high stakes or heavy judgement

    The work stays with a person for now, and the reason is recorded so that the rating can be revisited later.

  2. Assist

    Where most knowledge work sits today

    AI drafts and prepares the work, and a person reviews it and owns the result.

    Indicated

  3. Automate

    Typically high volume, clear rules and easy to check

    AI does the work, and people check samples and handle the exceptions.

Rating factors

Rate a workflow

Choose a setting for each factor. Low scores 1 point, Medium 2 and High 3. For Low Stakes, choose High if an error would cost little. For Low Judgement, choose High if the work needs little specialist judgement.

Volume

How often the workflow runs, and how much team time it takes.

Clear Rules

Whether a capable new starter could learn the workflow from a short walkthrough.

Easy to Check

How quickly someone can check the output, compared with doing the work themselves.

Low Stakes

How small the cost of an uncaught error is, and how sensitive the data involved is.

Low Judgement

How little the workflow depends on one person’s expertise or relationships.

Assist 10 of 15 points

Use AI to prepare the work. A person checks the output and takes responsibility for the result.

Review the result with the people who do the work before changing how the workflow runs.

What would justify investing beyond the pilot?

Agree on the baseline before introducing the tool. For the distributor, useful measures include the time to resolve each enquiry, whether the promised delivery date was accurate and how often the customer had to call again.

A reduction in handling time releases capacity. To estimate the financial return, management also needs to know how that capacity will be used, what the revised process will cost and whether service quality holds as volumes rise. Give one senior owner responsibility for that decision.

Two service advisers review a customer enquiry together at a computer.

  1. Measure the starting point

    Record volume, cost, turnaround and error rates before introducing AI.

  2. Run a measured pilot

    Agree the success threshold and stopping rule, then decide whether to scale or stop.

  3. Scale what works

    Expand one group at a time and track benefits in named lines of the profit and loss.

Phased benefit delivery from The AI Playbook for Mid-Market Leaders, page 34.