AI training for executives: from a proposal to a pilot decision

At Syntalith, we propose training for leadership teams covering process selection, alternatives for improving the work, and evaluation of vendor proposals. The program also addresses pilot preparation, responsibilities, and decisions to continue, change scope, or stop. Preparation starts with one initiative under consideration. We use it to select materials approved for training and tasks suited to each role, with feedback on participants' assessments and independent practice on a new case.

Define the work the company wants to change

“AI for customer service” is too broad a starting point for comparing proposals. It could mean finding information, drafting replies, or passing cases between teams. The program covers describing the selected part of the work: where delays arise, who handles it, and what must be ready before a case can move forward. The department manager brings knowledge of daily operations, while leadership explains why this problem deserves attention now.

That account provides a basis for considering alternatives. A feature in the current system may already support the task. Elsewhere, the answer may be team practice, connecting tools, or changing how information is passed on. Custom software needs to be considered alongside these options. During training, managers work on the reasoning behind a choice and the questions that need answers before commissioning work.

Read the proposal alongside the company's responsibilities

Another part of the program examines vendor scope. What does the prototype include, and what remains to be prepared for actual use? Who will provide access to sources and maintain connections to company systems? Participants compare the service description with the work their own team will retain. This brings dependencies into view that the vendor presentation does not show.

Reviewing results deserves separate attention. A manager needs to know who can assess an answer or document and how much attention corrections require. The number of generated texts does not describe that work. A pilot proposal therefore needs to compare the whole task with the current method, including gathering information and reviewing the result.

How much work remains after a response is generated?

In a hypothetical demo, AI drafts a polished service response. Before it can be used, the manager still needs to find the customer's facts and check whether the text promises something the company has not agreed to provide. A quickly written draft does not establish how long it takes to prepare a response ready to send.

A trial should also include finding the facts, reviewing commitments, and making corrections. Comparing this with the current process will show where the work has changed and whether the result meets the agreed requirements. If the team gains capacity, its intended use still needs to be defined; more available staff time does not automatically reduce spending.

A pilot that supports a next decision

Managers work on defining the trial: which cases it covers, who will review them, and what observations will be needed at the end. The criteria relate to the work the company wants to change. They may concern response quality, review effort, or whether the task can be handled in the company's existing environment. It is also useful to specify what the trial excludes, such as work performed by teams in other departments.

The discussion at the end of a pilot includes several possible decisions. Expanding the project requires confirmation that the team can handle the additional scope. A change may be needed when the tool works but the wrong part of the process was chosen. Stopping is a reasonable option if the solution does not meet the agreed requirements.

The 2023 NIST AI RMF 1.0 provides a useful reference. The voluntary framework covers governance, mapping context, measurement, and risk management. Assessment is revisited during a project as conditions change or new information emerges. NIST states that a revision is underway.

A shared decision, different responsibilities

Board members and executives need to weigh the initiative against other priorities and examine the case for continued funding. Department managers assess changes to team workflows, staff availability, and output quality. IT and procurement clarify technical dependencies, vendor scope, and ongoing operation. A shared part of the training brings these perspectives together, while role-specific tasks reflect each person's contribution to the decision.

Management experience does not determine familiarity with AI. Someone new to the tool needs to understand how a draft is prepared and checked. A regular user can concentrate on comparing proposals and evaluating assumptions. We agree on the introductory material, session format, and any practice between meetings before training begins.

To judge the usefulness of the sessions, look at the assessment participants prepare: can a reader understand the proposed trial, the company's responsibilities, and the basis for the next decision? Instructor feedback addresses gaps in those passages, and the author revises their own assessment. In a later independent task, information about access to required data changes, prompting a fresh assessment of the pilot's scope.

If the main difficulty concerns tools that have already been purchased but see little use, our article on increasing AI adoption after buying licenses addresses that situation.

An initial conversation with Syntalith only needs a description of one stalled AI decision. We will discuss what is holding it up, which parts of the program fit the situation, and who should take part. That gives us a basis for discussing materials and exercise preparation without requiring a confidential proposal at the outset. See our pricing page for information about working together.