When leadership decides that the organization needs an AI initiative, the next question often slows the conversation down: where should the company begin?
Sales may want help preparing customer information. Customer Service may want to reduce repetitive work. Operations may want faster handoffs, while IT has to consider systems, data, and security at the same time. If the discussion begins with a catalogue of AI technologies, the project can become too broad for anyone to define the intended result.
A stronger starting point is a business problem with a visible owner, an observable impact, and a process the team can examine. The organization can then decide where AI is relevant within that problem.
“We want AI” is an ambition, not yet a project brief
AI describes a possible approach. It does not explain where the business is struggling. Requests such as “we want an AI agent” or “we should use generative AI” show interest in a technology, but they do not provide enough information to design a solution.
A workable problem statement should identify:
• Who experiences the problem
• Where it occurs in the process
• How the current method creates effort, delay, inconsistency, or risk
• What a better outcome would look like
• Which data, systems, approvals, or controls affect the work
For example, “service agents spend time after each call preparing notes and routing the case” offers a clearer starting point than “we want Voice AI.” The first statement gives the team a process, a user group, and an operational outcome to investigate.
How to find a business problem that AI can address
1. Look for frequent work that consumes people’s time
Repeated tasks are relatively easy to observe. They may include reviewing documents, summarizing conversations, classifying information, or locating answers across several sources. Frequency alone does not make a task suitable for AI, though. The team also needs to understand the rules, exceptions, and information involved.
2. Find information that exists but remains difficult to use
Enterprises often hold valuable information in recordings, documents, messages, and separate systems. The problem may be accessibility rather than availability. Information can be difficult to search, poorly structured, or unavailable to the person making a decision. AI may help process, organize, or retrieve it, depending on the use case.
3. Examine decisions that require extensive preparation
Some tasks do not repeat in exactly the same form, yet employees have to gather and compare information before making a decision. AI may assist by preparing the relevant material or highlighting items for review. The final judgment can remain with a person when the process carries significant risk.
4. Learn how frontline users actually work
A documented workflow rarely captures every workaround. Employees may switch between screens, copy information manually, or keep personal files to bridge gaps between systems. Observing and interviewing users reveals pain points that process documents often miss.
A practical way to prioritize AI use cases
Once the organization has a list of problems, it does not need to address all of them at once. Five dimensions can make prioritization more disciplined.
Business impact
How does the problem affect revenue, cost, service quality, risk, or customer experience? Use evidence already available in the organization rather than relying on broad promises of “greater efficiency.”
Process clarity
Does the team understand the current workflow, including handoffs and bottlenecks? If the process remains unclear, adding AI may transfer the existing confusion into a new system.
Data readiness
What information does the use case require? Where is it stored, what form does it take, and can it be used under the organization’s policies? Early inspection makes the feasibility assessment more realistic.
Integration readiness
Which systems need to provide or receive information? If people must copy every result into another application, the solution may improve only one part of the job. Mapping the data flow and system owners early helps expose hidden dependencies.
Measurability
A first use case should have a reasonable baseline, such as the time required, number of steps, rework volume, or an internal quality measure. The metric should match the problem. It does not have to begin with a dramatic financial target.
Scenario: Turning “use AI in the contact center” into a workable brief
Imagine that a service team proposes using AI in its contact center. That request could include answering questions, transcribing calls, analyzing conversations, or supporting quality reviews. The scope is too broad to design responsibly.
After observing the workflow, the project team might discover that agents must summarize each conversation, select a case type, and route the information to another team. A more focused brief could become: “Prepare the summary and required handoff information, with an agent reviewing it before submission.”
This version identifies the user, input, review point, and expected output. The organization can design a controlled test, evaluate the output, and identify where integration with the existing system is required without attempting to build every capability at once.
Start small without losing the end-to-end view
Choosing one initial use case does not mean planning in isolation. The organization should still understand the full path across data, users, systems, controls, and post-launch ownership. It can then select a first scope that provides meaningful learning with manageable risk.
A practical sequence is:
1. Define the problem and its owner
2. Map the current workflow based on real work
3. Establish the intended outcome and baseline
4. Review data, systems, controls, and human checkpoints
5. Test within a manageable scope
6. Collect user feedback and improve the solution
7. Decide whether to expand based on evidence
Begin with a problem that matters and can be examined
An enterprise does not need a strategy that answers every possible AI question before it begins. It does need a clearly defined problem, engaged process owners, inspectable data, and a way to determine whether the work has improved.
AppMan’s One Partner approach connects Business, Operations, and Technology across problem discovery, planning, solution design, development, integration, and ongoing optimization.
If your teams have several AI ideas but no clear starting point, contact AppMan to assess the business problems, prioritize use cases, and develop a Success Blueprint for your enterprise AI initiative.
TH