PUT AI TO WORK

40% of AI projects may be canceled. Start smaller.

The businesses that get results start smaller, measure earlier, and choose tools they can afford to run.

Issue #9 · August 19, 2026

AI is moving from experiment to everyday business tool. But the companies getting value are not starting with the biggest project. They are starting with the clearest one.

This issue contains affiliate links. If you purchase through these links we may earn a commission at no additional cost to you.

Start with a smaller AI project

Gartner says 40% of agentic AI projects will be canceled by the end of 2027. The reasons are familiar: costs rise faster than expected, implementation turns out to be more complex than planned, or the use case was never a good fit for an agent in the first place.

That statistic is not a reason to wait. It is a reason to make better decisions before day one. And you do not need an enterprise budget to do that. In fact, 57% of U.S. small businesses are now investing in AI technology. This is no longer early-adopter territory. The question is whether you deploy well.

Start with a bounded, high-volume task. The strongest first use case is usually not the most ambitious one. It is a task with a clear beginning and end that happens often and follows a repeatable pattern: answering common support questions, sorting incoming requests, or moving order details from one system to another.

A bounded task gives you a clean test. You can see where the agent helps, where it needs a person, and whether the output is good enough to use. A broad goal like "automate customer service" is too hard to measure. "Draft answers to our 20 most common questions and route exceptions to a person" is much easier to start.

Define success in numbers before launch. Decide what a good result looks like before the first task runs. Is the target a 30% reduction in response time? A 70% first-pass resolution rate? Five hours saved each week?

Without a target, every outcome feels ambiguous. A tool can appear impressive while saving no meaningful time, or it can look imperfect while still producing a worthwhile return. Your starting numbers turn a vague experiment into a decision you can make.

Match the tool cost to a realistic budget. Small projects often fail when the price expands along with usage. Before you choose a tool, estimate your monthly volume, likely plan level, and the cost of the human review still required. Leave room for the project to grow, but do not pay for growth you have not earned yet.

Your first AI project does not need to transform the whole business. It needs to solve one recurring problem well enough that you can prove the value and decide what comes next.

AI compliance isn't coming – it arrived August 2nd.

The EU AI Act transparency deadline passed on August 2, 2026. If your business serves customers in the European Union and uses AI in customer-facing interactions, this is now a practical check for this week–not a future planning exercise.

You do not need to start with a long legal memo. Start with three questions about the tools you already use.

1. Can customers tell when they are interacting with AI? Review chat widgets, support assistants, email responders, and other customer-facing features. If an AI system is involved, check whether the experience clearly discloses that fact. Ask your vendor where that disclosure appears and whether you can control its wording.

2. Do you know whether any tools fall under a high-risk category? The Act's Annex III lists high-risk use cases. Make a simple list of the AI tools your business uses, what each one does, and whether it influences a decision in an area covered by those definitions. If you are unsure, raise the question with the vendor instead of assuming the answer.

3. Have you reviewed your vendor's compliance documentation? Ask for the vendor's current AI Act materials, product disclosures, and stated responsibilities. Third-party tools do not remove the need for you to understand how their features affect your customer experience. Save the answer somewhere your team can find it when a product changes.

This is not only a European issue. Colorado and California have operational state frameworks, and the FTC continues to scrutinize claims about what AI products can do. The practical move is simple: make a list of customer-facing AI tools, check their disclosure features, identify anything that may be high-risk, and request updated compliance information from each vendor.

The fastest AI ROI most small businesses can get – by the numbers.

Customer service is one of the clearest places to test an AI agent because the work is frequent, bounded, and repetitive. The numbers in the brief are compelling: 92% of businesses that deployed AI customer service report improved satisfaction scores, with an average cost reduction of 52%, 40% more inquiries handled, and resolution time falling from 8.2 hours to 1.3 minutes.

Those figures are not a promise for every business. They are a reason to check whether your operation has the right shape for a useful first test. If your team handles more than 20 support inquiries per week manually, you are a candidate.

Start by measuring three things for one week: the number of incoming requests, your current average response time, and your current cost per inquiry. Then separate the requests into common questions, requests that need account-specific information, and exceptions that should go straight to a person.

That gives you a sensible first workflow. The agent can handle the common questions, pull approved information into a draft response, and route anything unusual to your team. Tools like Make.com can connect your inbox, CRM, and response agent into a single workflow, so a new inquiry does not have to be copied manually between systems.

Keep the first version narrow. Give it a defined set of questions, a clear handoff point, and a way to measure what happened. After two weeks, compare response time, cost per inquiry, and the percentage of requests resolved without escalation with your original baseline.

If the numbers move in the right direction, you have a case for expanding. If they do not, you have useful information before committing more time or money.