What I’ve learned about business problems.

Small firms are buying AI for writing and marketing. Almost nobody is pointing it at a process, and the reason is not money or skills.

Across five unrelated engagements the same shape keeps appearing: the return went up when the amount of activity went down.

A low cost per lead is a price. Ten phone calls will tell you which of your sources is actually buying you customers.

DOMIK UA had the visitor numbers a real estate platform is supposed to want. The developers paying for leads weren't renewing, and the traffic dashboard had no way to say why.

The hardest part of automation is not the tool. It is choosing the work. Six questions rank every repetitive task in a business by what removing it is worth, and they usually put a dull job at the top and the exciting one near the bottom.

The usual advice is to automate everything you can. The cases where AI went wrong in public point to a narrower rule: automate the work where a mistake is cheap to catch, and keep a person on the work where a mistake becomes a promise, a payment or a decision about someone.

Agents are the most sold and least understood thing in AI right now. Most of the work a small business wants to automate needs rules, not decisions, and one 2025 benchmark of office tasks shows why that matters.

Most AI purchases go wrong in the contract and the first month, not in the demo. Seven steps, each with a test you can run yourself, cover what the tool removes, where your data goes, who owns what it produces, and how you get out.

An AI pilot without a baseline cannot fail, which means it cannot succeed either. Six numbers, agreed before the tool goes in, turn a pilot from a demo that never ends into a decision.

Most AI tools sold to small businesses promise to score leads, write follow-ups, forecast sales or answer customers. Sales and service AI tools depend heavily on the customer records, conversations and outcomes the business already keeps. Where those records live in inboxes, spreadsheets and someone's memory, the AI has nothing reliable to work with. The order that works is CRM first, then AI on top of it.

The two jobs owners most want to hand to AI are answering enquiries and following up leads. They are also where slow replies cost the most sales. The right setup is mostly plain automation with rules, AI for the messy middle, and a person wherever the reply commits the business. Here is how to build it in that order.

An AI automation agency builds automations and is paid to build more. An AI consultant is supposed to tell you what is worth automating, including when the answer is nothing. Many firms call themselves both. The difference matters most at the start, when the question is not how to build but what to build.

A startup's first ten customers almost never arrive through marketing. They come from one buyer narrow enough to list by name, a run of direct conversations, a commitment that costs the buyer something, and delivery done by hand. Here is the order I run it in, and the numbers that tell you it is working.

A product with no sales has no price history, no competitor that matches it exactly and no data on what customers will pay. Most founders fill that gap with a cost-plus guess or a competitor's number. There is a better order: find out what the buyer pays today, decide what the price is charged on, and test it with money rather than opinions.

A B2B startup without a sales team does not need a go-to-market deck. It needs four decisions made in the right order: who buys first, how they find out, what the page and the first conversation have to prove, and how the founder's way of selling gets written down so someone else can repeat it.

A deck earns the meeting. What earns the money is what an investor finds after it: who the founders are when the slides are closed, whether customers pay, what a customer costs, and whether the plan for the money survives a few hard questions. Here are the six checks, and what a startup can prepare for each before anyone asks.

A working prototype proves the product can exist. The first production run proves the company can make it again, on a date, at a cost. Between the two sit purchasing lead times, part choices made for the bench rather than the line, and a factory chosen before anyone knew the volume. Here is what to settle before you place the first order.

Founders use advisor, mentor, consultant and fractional executive almost interchangeably, and then are surprised by what they get. The four do different jobs, carry different responsibility and are paid in different ways. Here is how to tell them apart, what advisor equity is standard, and how to choose by the problem you have rather than the title you like.

Most startups build their pricing page by copying a competitor's three columns and filling them with features. Packaging works better from the other end: what a customer gets value from, which kinds of customer you serve, and what makes one of them upgrade. The tiers follow from those answers, not from the feature list.

Three manufacturer quotes arrive in three formats, on three sets of terms, and the lowest unit price is usually the one that costs most by the time the goods are on your shelf. Here is how I put quotes on one basis, the seven lines that move the real cost, and the one question that tells you most about a supplier before you order.

The deposit is the moment a supplier relationship becomes expensive to leave. Before it, you can walk away at the cost of a few emails; after it, every problem is a negotiation. Here are the eight checks I run before a client pays a new manufacturer anything, and the two that catch most of the trouble.

A supplier audit before the first production run is not a tour of the factory. It is a check of whether the process that made your approved sample will make the next thousand the same way. Here are the eight areas I look at, the certificate check that takes five minutes, and the one document that tells you most.

Choosing a supplier gets weeks of attention. Running one gets almost none, until an order is late and nobody can say whether it is the first time or the fifth. A supplier scorecard fixes that with six measures a small company can keep in one sheet, reviewed once a quarter, and used in the next price conversation.

The delivery date you promise a customer is set by the slowest thing you buy, plus every day an order waits on your own side before it reaches the supplier. Most lead-time projects attack the supplier first. The faster wins are usually in your own handoffs, in the order you buy things, and in choosing parts that are made in larger numbers.

At a few hundred orders a month, a third-party logistics provider can take packing, storage and shipping off your desk, or add a monthly invoice you do not understand and a support queue you cannot reach. The difference is decided before you sign. Here are the seven questions I would ask, and the test that tells you most.

Most deals between a company and its supplier or partner do not collapse. They stop. Emails slow down, calls get moved, and both sides wait for the other to say the thing neither wants to say. Here is how I restart a stalled deal, why a neutral person in the middle works so often, and when it is time for formal mediation or counsel instead.

Most people looking for a manufacturer start with a directory and the first factory that answers. The search goes better in a fixed order: a broad survey of who can make it, prices at several volumes together with how reliably each one delivers, and a hard check of quality. Then the choice of country, which is about total cost, time and control, not the unit price.

A minimum order quantity looks like a wall. It is usually a cost the supplier is protecting: a machine setup, a raw-material batch, a production slot. Once you know which one, you can trade for it, with a higher price, a committed schedule, a standard material or a split delivery, instead of asking for an exception the supplier has no reason to give.

A purchase order tells a manufacturer what to make this time. A contract manufacturing agreement decides what happens every other time: when quality slips, when a delivery is late, when prices rise, when you want to leave. Most problems with manufacturers are easier to solve because a term was settled in advance. Here are the twelve to settle, in plain language for the commercial side.

A small company with one supplier for a critical part is one fire, one strike or one customs hold away from stopping. A second source fixes that, but it costs money and attention, and done badly it annoys the main supplier without giving you a real alternative. Here is which parts to dual-source first, how to split the volume, and what to do when a delay hits anyway.

The price of a small business is usually a multiple of one number: seller discretionary earnings. Every dollar of SDE you accept is worth two or three dollars of price, so every add-back the seller claims deserves the same scrutiny as a price increase. Here is the order I check it in, and the question that matters more than the total.

In many small businesses, the owner is the sales team, the senior technician and the reason the best customers stay. Buy the business and the owner leaves. The risk is not that dependence exists; it is that nobody measured it before the price was set. Here are the seven places to look, and how each one should change the deal.

Most checklists for buying a business stop at the warning sign. The useful question is what to do about it. A red flag is rarely a reason to walk away; it is a reason to change the price, hold part of it back, pay part of it later, extend the handover or add a condition. Here are nine I see most often, and the term each one should move.

Owners preparing to sell usually tidy: repaint, clean up the website, chase a few overdue invoices. Buyers do not pay for tidiness. They pay for earnings they can verify and believe will continue after the owner leaves. Preparing a business for sale means building that proof, and it takes a year more often than a month.

A buyer pays for earnings that will still be there after you leave. If the customers, the difficult work and the decisions all run through you, a buyer either pays less, pays later or walks away. Taking yourself out takes a year more often than a month. Here is the order I would do it in, and how to prove to a buyer that it worked.

Agency reports are full of numbers that go up: impressions, clicks, engagement. The number an owner needs, money spent against customers won, is usually missing. Here are seven checks I run when an owner suspects the spend is not working, starting with the one that decides most: can anyone trace a single lead from the ad to the sale?

Buyers spend months on the price and minutes on the cash the business needs the day after closing. Then payroll, a supplier invoice and a slow-paying customer arrive in the same week. A profitable business can still run short of cash if what transfers at closing does not match its cash cycle. Here is how to work out what you need, what should transfer with the business, and how much cushion to keep.

Most small businesses sell for a multiple of their seller's discretionary earnings, and the multiple depends heavily on size. In the second quarter of 2026, brokers reported about 2.0 times earnings for the smallest deals and about 3.1 times for those between $1 million and $2 million. The number matters less than what moves a particular business up or down the range.

A business that earns a large share of its revenue from one customer can look excellent until that customer leaves. Buyers ask for a percentage that is "too much", and there is no official one. What there is: an accounting line at 10% that public companies must disclose, and a set of questions that decide whether a large customer is a risk or an asset.

The phone rings, the forms come in, and the month ends with fewer sales than the leads promised. Owners usually blame lead quality and buy more leads. Most of the time the leads are fine and the break is after them: how fast someone answers, what happens on the first call, and what happens when the prospect goes quiet. Here are the five places to look, in the order I check them.

When leads are poor, the instinct is to buy more of them and hope the good ones are in there somewhere. That multiplies the cost of sorting. Lead quality is mostly decided before the lead arrives: by what the ad promises, who it reaches and how much effort the form asks for. Here is the order I fix it in.

The jobs wait for your answer, the team waits for your approval, and you are still fixing things at eleven at night. Most owners try to fix it by hiring, and the new person waits for approval too. The bottleneck is not your workload. It is the rules that live only in your head. Here is how to move them out, one decision at a time.

When profit is thin, owners argue with themselves about raising prices. Sometimes the price is the problem. Often it is the offer around the price, or the sales that are lost before price is ever discussed. Here is how I tell the three apart, why price is the strongest lever when it is the right one, and how to raise it without a wave of cancellations.

Sales have been flat for a year, the team is busy, and nobody can say why the numbers will not move. Owners usually spend on the part they understand best, often more marketing. A business stalls in one of five places, and only one of them is fixed by more leads. Here is how to find which one is yours before spending anything.

Owners who need help with marketing usually choose by budget: an agency if they have some, a freelancer if they have less, a hire if they have more. The better question is which problem you have. Doing the work, deciding what work to do, and checking whether it worked are three different jobs, and each option covers a different one.

Consultant prices vary more than almost any other service a small business buys, from free volunteer mentoring to fees that rival a senior salary. The number on the quote matters less than what it is charged for and what it should change. Here is how consultants charge, what drives the price, where free help exists, and how to judge a quote by the result it promises.

Most owners who want to grow think about new leads. One of the most common reasons growth stops is the opposite: customers buy once and never come back, so every month the business pays to replace the ones it lost. Repeat buying is cheaper to win, easier to predict and usually where the profit is. Here is how to measure it and raise it.

The accountant says the business made a profit. The bank account says otherwise. Profit and cash are different measures, and a business can show one while running short of the other. The gap usually sits in one of five places, and each has a different fix. Here is how to find yours in an afternoon.

Most owners cannot say whether their marketing pays. The reports show clicks and reach; the bank shows spending; nothing joins the two. Measuring return does not need an analytics team. It needs every customer tied to where they came from, profit rather than revenue, and the value of a customer over time rather than the first sale.

The title covers everyone from a retired executive giving free advice to a firm that rebuilds the whole operation. What a good one does is narrower: finds where the business is losing money or stalling, often somewhere other than where the owner looks, decides what has to change, and helps put the change in place until it works. Here is when that is worth paying for, and when it is not.

The three options are often treated as price points for the same job. They are not. A full-time COO builds and runs the operation for years; an interim COO fills a gap full-time for months; a fractional COO holds the plan and the numbers part-time while the company grows into the role. Choosing the wrong one costs more than the fee. Here is how to tell which job you have.

If the role cannot change any decision, it is closer to consulting than operating. If it decides everything, it is a CEO you did not hire. The role works when the line between them is drawn in writing before the first meeting. Here is what the seat does every week, the decisions that usually move to it, and the ones that never should.

A fractional COO is the right answer for a specific problem: a growing company where every cross-department decision lands on the CEO and nobody senior holds the plan between meetings. For several other problems that look similar, it is the wrong answer, and an honest fractional executive will say so. Here are the signs for, and the five cases against.

The offsite ends with energy, a whiteboard of priorities and a deck. Three months later, most of it has quietly stopped. Strategy does not fail in the room. It fails in the weeks after, between meetings, when nobody owns the next step. Here is how I turn an offsite into a 90-day operating plan that is still moving at the end of the quarter.

Most leadership meetings are status updates read aloud. Everyone reports, the hour runs out, and the decisions that were needed are pushed to next week, where they are discussed again. A weekly leadership meeting should do one thing: decide what the numbers say needs deciding, and write it down. Here is the structure I use.

Most growing companies have either too few numbers, so the CEO runs on instinct, or too many, so nobody looks at any of them. A company of 50 to 500 people can be run on about twelve, if each one has a single named owner who explains it every week. Here are the twelve I start from, and how to make them work.

The three titles get used for the same idea, a senior person part-time, but they own different things. A fractional CEO leads the company; a fractional COO makes the company's machinery work for the strategy; an operating partner works across a portfolio for an investor. Many owner-led companies looking for a second CEO are really looking for operational ownership.

Most fractional COO engagements that fail were set up badly: no written brief, a candidate chosen on chemistry, no decision rights, and no review date. Hiring well takes a page of brief, one real problem as a test, a written list of what the seat may decide, and a review after ninety days. Here is the order.

Most strategy sessions produce energy and a whiteboard, and three months later almost nothing has changed. The session is not where strategy fails; the weeks after are. A session designed for what happens afterwards prepares the numbers in advance, spends time on what to stop, and ends with three priorities, named owners and a weekly check already in the calendar.

A program is funded, the people it is for exist, and too few of them apply. Low enrolment gets read as low interest. The research says otherwise: people miss out because they never heard, could not tell whether they qualified, found the process forbidding or feared what applying would cost them. Each cause has a different fix, and fixing the first barrier is often cheaper than sending more people into a process that still stops them.

The data is late, the sources contradict each other, and the decision date is fixed. Waiting for clean numbers is not an option, because the decision will be made anyway, with or without evidence. The method is to find out what each source is really counting, combine what can be combined, and keep each source's limits attached to the conclusion.

When aid or services run through dozens of volunteer groups, partners and regional offices, the same request gets answered twice while another is missed. The usual fixes are a central hub that approves everything, or a new platform. Both fail under pressure. What works is a written answer to one question, what can a local team decide alone, plus a few light formats people use without being reminded.

A study takes months, runs to two hundred pages, and reaches the people who must decide with the answer buried on page 140. They read the summary, which describes the study rather than answering the question, and decide on instinct. Research gets used when it is written backwards from the decision. Here is the structure that works.