AI Implementation for Business Owners: Start With the Business, Not the Tool

You sign up for an AI tool, connect a few accounts, and ask it to help run your business. The answer looks polished. It also misses details, relies on old information, and leaves you wondering whether you have done something wrong. Meanwhile, someone online claims to have automated their entire company over the weekend. Suddenly, experimenting with new technology feels like falling behind.

That frustration runs through Aaron Spivak’s conversation with Mo from Zero to Agent on the Founder2Founder podcast. With a background in events, advertising, and now AI implementation, Mo discusses what he sees inside businesses trying to adopt these tools. The episode covers agents, data, employee concerns, and advertising. Its most useful message is that meaningful implementation starts with understanding the business you already have.

Start With the Result You Want

When businesses approach Mo, he says their priorities usually involve increasing revenue or reducing expenses. Those are goals, but they need more detail before they can guide useful work. A business struggling to follow up with enquiries has a different problem from one producing advertisements too slowly. Both may benefit from AI, but they do not need identical solutions. The starting point is the outcome the owner wants to improve.

Spivak challenges the expectation that every business has an obvious opportunity to become several times more effective overnight. Some companies have extensive business information that support particular use cases. Others may initially find value in smaller tasks or individual processes. The usefulness depends on the work involved. An example from another business does not automatically describe your own opportunity.

This is why Mo frames adoption as a business question before a technical one. Leaders need to identify where information gets lost, where work slows down, and what a better result would look like. A technical team can help build a solution. It still needs a clear problem to solve. Choosing tools without that direction can create activity without resolving the underlying difficulty.

Understand What You Mean by an AI Agent

Mo separates the discussion into three broad stages: using models directly, building automated workflows, and creating agentic systems. In the first stage, someone asks a tool to draft, summarise, or analyse information. Workflows connect steps that would otherwise require manual effort. The third stage involves systems designed to coordinate tasks across a broader set of tools and instructions. These distinctions help explain why two businesses can both say they use AI while doing very different things.

He introduces the term chief agent officer to describe a coordinating AI layer. In his proposed setup, it receives requests and delegates work to more specialised agents. He compares it to a chief of staff supporting different business functions. That is his model for organising the work, rather than a requirement every company has already adopted. Its relevance depends on what the business needs coordinated.

The conversation later also considers appointing a person to lead AI adoption. That human role should not be confused with the coordinating agent Mo describes earlier. Someone still needs to understand the business, decide what to implement, and remain accountable for the results. Giving a system a leadership title does not remove those responsibilities. The organisation needs clarity about both the technology and its human owner.

Your Data May Be the Real Starting Point

Spivak argues that AI often exposes how disorganised business information has become. Documents sit in different folders, conversations contain missing context, and important details remain inside separate platforms. Connecting those sources can produce an answer without producing a reliable picture. The tool may be working with incomplete or inconsistent information. That makes it difficult to know how much confidence to place in its response.

Mo shares an example of a company eager to explore advanced AI applications despite lacking a central customer relationship management system. Documentation lived in email, and executives received reports through separate messages. Before discussing sophisticated agents, his team needed to understand that setup. The gap was organisational as much as technical. There was no clear starting point for finding the information the system would need.

His assessment begins by locating data and understanding its intended use. Spivak adds that labelled folders alone may not be enough if the relevant information remains inaccessible to the tools involved. The broader question is whether a business can identify information it trusts and keep it usable. Better access does not automatically mean better quality. Both need attention before outputs become the basis for important decisions.

The Founder’s Excitement May Not Match the Team’s

Mo says his assessments also look at employees’ readiness to work with AI. In the companies he describes, founders often express greater enthusiasm than the people expected to use the new systems daily. He sees that difference as an important part of implementation. A founder may attend a few planning meetings and feel ready to proceed. The team still has to change how the work gets done.

Spivak points out why employees may feel uneasy. They are expected to maintain performance, learn unfamiliar tools, and think about what automation could mean for their jobs. Even capable people may find those demands difficult to manage together. Enthusiasm cannot simply be assumed because leadership sees an opportunity. The practical and emotional sides of the change arrive at the same time.

Mo’s response is to include the team in the assessment rather than focus only on the founder. Understanding their confidence, knowledge, and concerns helps shape the next steps. It also reveals where support or explanation is needed. The people doing the work often understand details that an ambitious plan overlooks. Their participation matters to whether a new process becomes part of everyday operations.

Consider Capacity Alongside Cost Savings

One of the episode’s most useful examples involves a business with roughly ninety editors. Mo describes discussing an arrangement linked to savings from AI implementation. The founder saw another possibility: keeping the team and using the technology to handle more work. There was already demand the business wanted to serve. Additional capacity could therefore be valuable without reducing staff.

That response stood out to Mo because the founder had already invested in training those people. They understood the work and were part of a team he valued. Helping them produce more could support growth. In that situation, the question was about how much work the existing business could take on. Payroll savings were only one possible way to evaluate the opportunity.

The example gives business owners a broader question to ask. What would happen if the team could spend less time on repeated tasks and more time delivering useful work? The answer might involve faster service, additional projects, or more room for complex tasks. It will depend on demand and the business model. The important point is to examine the available capacity before assuming that efficiency must mean fewer people.

Use Advertising Data to Ask Better Questions

Mo’s advertising background shapes much of the practical discussion. He describes using AI to examine creative assets alongside performance information. A team might compare different openings, visual styles, messages, and product presentations across a large collection of advertisements. That can help organise observations that would be time consuming to review manually. The goal is to generate more informed ideas for what to test next.

He also discusses creating variations around concepts that have already attracted delivery or produced results. Instead of beginning every creative cycle from scratch, the team can use previous work as a starting point. New versions still need assessment. Producing more assets does not, by itself, demonstrate that the advertising is improving. The useful part is connecting production with feedback from actual campaigns.

Spivak shares an anecdote about a member whose AI-created advertisement performed strongly compared with other creative. The story illustrates why founders are interested in experimenting. It is one reported outcome, rather than a promise about what any new advertisement will achieve. The broader opportunity discussed is faster iteration. A business can explore more ideas while learning which ones connect with its customers.

Read Performance Through the Whole Customer Journey

When discussing campaign analysis, Mo uses the familiar sequence of attention, interest, desire, and action. Attention concerns whether someone stops to look. Interest considers whether they continue watching. Desire looks at whether the message encourages a click. Action brings the discussion to conversions and what happens after the initial response.

That structure helps organise the different questions an advertisement raises. A strong opening may attract viewers who never continue to the offer. Another creative may generate clicks while the page fails to persuade visitors to buy. Looking across the sequence gives the team more context than focusing on one impressive number. It also creates more specific questions for an AI-assisted review.

Mo places considerable emphasis on how much spend the platform allocates to an advertisement. That is his preferred signal within the approach he describes. It should be understood separately from proof that an advertisement is profitable. The business still needs to examine the outcome it is paying to obtain. The useful takeaway is to connect creative observations with results rather than treat production volume as the achievement.

Give AI Access With a Clear Purpose

The conversation returns to access and security near the end. Mo says businesses regularly ask what information they can share and how to connect systems appropriately. His team helps define boundaries around how tools operate. This matters particularly when a proposed system reaches across several parts of a company. More connected information also means more decisions about who or what should be able to use it.

One example he calls a founder command brief brings relevant information together to suggest priorities. It may also help prepare responses based on the context available. The intended benefit is to reduce the time a founder spends moving between scattered updates. Instead of reacting to every message separately, the owner gets a more organised starting point. The quality of that brief still depends on its sources and instructions.

This returns the discussion to business ownership. Someone must decide which information belongs in the brief and what deserves attention. Someone must also decide when a draft or recommendation needs review. The technology can support that process, but a useful implementation needs clear expectations. Connecting everything and asking an open question is a different approach from designing support for a defined responsibility.

The Founder Still Defines Success

Spivak offers a comparison between two businesses starting at one million dollars in annual revenue. One aims to double and finishes at two and a half million. The other targets ten million and reaches seven million. The second has grown more, yet its team may feel it has failed. The first may feel proud, satisfied, and confident about what it achieved.

His point is that the target shapes how people experience the year. Working hours, financial pressure, family time, and expectations all influence whether the result feels worthwhile. A larger revenue figure does not answer every question about the business or the owner’s life. Leaders choose what they are asking their teams to pursue. That decision matters before selecting the tools intended to help them get there.

AI cannot settle that personal question simply by producing a more ambitious plan. The founder still needs to decide what kind of company they want to build. More capacity might support rapid expansion, or it might create room for a more manageable working week. Both possibilities require deliberate choices. The technology becomes useful in relation to those choices rather than supplying a definition of success on its own.

Move Forward Without Chasing Every Headline

Spivak closes with a call for active experimentation alongside patience about the bigger picture. He questions dramatic claims about tiny companies and sweeping workforce changes rather than accepting every headline at face value. That attitude fits the most practical parts of the episode. Business owners can explore new possibilities while still asking what the evidence shows. Curiosity does not require abandoning judgement.

The next step therefore does not have to be rebuilding the whole company around agents. It can begin with a clear business problem, an honest look at the available data, and a conversation with the people doing the work. From there, the owner can decide what support is worth building and how to assess it. The aim is a business that works better on its own terms. That is more useful than collecting tools simply to feel caught up.

‍