Getting Started With AI | Blane Canada

Question 1

Where should we start with AI?

Accepted Answer

Start with a problem, not a tool. Before buying anything, run an operational audit to find where time and money are actually being lost, set basic guardrails for how AI can be used, and train your people on how the tools work. Audit, guardrails, and training form the foundation everything else is built on.

Most companies start by shopping. They follow a consultant's recommendation, chase a developer's demo, or go looking for the mythical killer app they read about in the trade press. That sequence produces activity without results, because adding technology to a process nobody has examined is a prescription for chaos.

The simplest and most direct entry point is three interconnected steps. First, an operational audit, which tells you where your time, money, and effort are actually going. Second, guardrails, which define what is acceptable use, what data is off limits, and who is accountable for output. Third, training, which builds the shared language and skill your team needs to use any tool well. Those three together create an AI strategy. Any one of them alone creates a project.

This approach also minimizes what you spend up front. You are not committing capital to a platform before you know which problem it is supposed to solve. You are spending a small amount of time to understand your own operation better than the vendors pitching you do.

Whatever you do, do not start by asking "where can we use AI?" Start by asking "where are we losing time, money, and decision quality right now?" Those are not the same question, and the gap between them is where most AI budgets disappear.

Question 2

What question should we ask before choosing AI tools?

Accepted Answer

Ask "where are we losing time, money, and decision quality right now?" instead of "which AI tools should we implement?" The first question produces a diagnosis. The second produces a purchase. When the solution comes before the diagnosis, the implementation solves the wrong problem very efficiently.

In four decades of watching organizations absorb major operational shifts, the pattern is consistent. When the solution comes before the diagnosis, the implementation solves the wrong problem efficiently. The waste stays invisible because the system is running and something is clearly happening. Dashboards move. Nobody stops to ask whether the right problem got solved, because the evidence of activity looks close enough to evidence of progress.

The second question is also harder to answer, which is part of why it gets skipped. Naming a tool takes an afternoon. Working out where your operation actually loses time and money means talking to people at several levels and comparing what they say to what the process documentation claims. That is uncomfortable work and it is where the value is.

It is worth noticing what the first question quietly assumes. Asking which tools to implement takes for granted that a tool is the answer, that the problem has already been identified correctly, and that someone has established what a good outcome would even look like. Three assumptions, none of them tested, all of them embedded in the question before anyone opens a vendor conversation.

The companies getting real traction are not the fastest adopters. They are the most deliberate ones. They spent time understanding where the operational drag actually lives before they spent money on anything.

Find the problem first. The tools come after.

Question 3

Should we start with an AI task force?

Accepted Answer

A task force is the right instinct in the wrong sequence. You cannot build an implementation plan by committee without a diagnostic first. Before guidelines and experimentation, you need an honest baseline: what tools are already in use, where the real capability gaps are, and what your governance exposure actually looks like.

A professional services firm recently sent an all-staff email announcing that a task force had formed and that guidelines, resources, and training were coming. More details soon. What struck me was that when the memo went out, the tools were already in the building.

The firm acknowledged something important. Their staff ranged from people with no interest in AI to people already building web applications with it. The team was dividing into adapters and non-adapters. That range is the real problem, and a task force memo does not close that gap.

Before the guidelines and the experimentation, the firm needed a baseline. An honest picture of what tools are already in use, where the capability gaps actually are, what is being considered, and what the governance exposure looks like in practice versus what leadership assumes. Those are three different things, and they rarely match.

The task force is the right instinct. It has a role. The sequence is wrong.

Question 4

Should our first AI project be large or small?

Accepted Answer

Small, narrow, and measurable. Pick one workflow. Fix the data issues in that workflow. Learn from it. Then expand. Small focused wins beat big vague moonshots, because a narrow project produces evidence you can act on and a broad one produces a status update.

Start with focused pilots aimed at narrow, well-defined problems. Gather information. Evaluate. Adjust. Expand. Micro-stack applications, meaning small contained builds that touch one workflow and one set of data, are the best place for most companies to begin.

The distinction that matters is not small versus ambitious. It is narrow and measurable versus broad and performative. A company that mandates "implement AI" and lets teams pick tools first ends up with press releases instead of results. One year later there is nothing to show, and the skeptics inside the building have won by default.

There is a second reason to start narrow. A contained project is where your team learns the unglamorous disciplines that determine whether anything larger will work: data quality, cleaning, validation, security, and ownership. Learning those on a small internal project is far cheaper than learning them on a customer-facing one.

Question 5

What is a good first AI project for a mid-sized professional services firm?

Accepted Answer

Start inside HR. Build an AI-assisted employee resource center that answers routine questions about benefits, pay dates, holiday schedules, and policies. It gives an understaffed function immediate relief, lets every employee experience AI as a helper rather than a threat, and teaches your team data discipline on a low-risk project.

If you cannot help your own employees find the holiday schedule without emailing HR, you are not ready for customer-facing AI.

HR is the right launchpad for three reasons. First, immediate return. HR is chronically understaffed in most mid-sized firms, and the volume of repetitive questions it absorbs is enormous. Second, cultural effect. After 40 years in business, I have learned that a change works best when it touches everyone positively. When an employee's first encounter with company AI is a tool that answers a question they were embarrassed to ask a person, it stops being a threat. Third, foundation building. Your HR content lives in PDFs, Word documents, spreadsheets, and outdated formats. Cleaning and structuring it teaches your team data cleaning, validation, and security on a project where the cost of a mistake is low.

You are not just deploying a chatbot. You are building the bridge to everything that comes after it.

Question 6

How do I keep up with AI without it becoming a second job?

Accepted Answer

Do not try. Build a filter instead. Ask whether a development changes one of five things: how your customers behave, how work gets done, what competitors can now do, how decisions get made, or what risk you carry. If it changes none of those, it is interesting rather than important.

I will say the uncomfortable part first. Even working in this full time, I have struggled to keep up. Success stories and cautionary tales pile up faster than anyone can read them. What I had to do was get ruthless about what actually matters to my clients and to our own work, and deliberately narrow what I pay attention to. If someone tells you they are on top of all of it, they are describing an ambition rather than a practice.

The five-part filter does most of the work. A new model with a better benchmark score usually changes none of those five things for a mid-sized manufacturer. A pricing change that makes a previously uneconomical process suddenly viable changes all of them. One deserves a meeting. The other deserves a bookmark.

The second half is people rather than feeds. A small number of trusted sources who can interpret technology through a business lens will save you more time than any newsletter. What you want from them is not news. It is the judgment call about which piece of news required a decision from someone in your position.

The goal is not knowing everything. It is knowing what matters soon enough to respond intelligently, which is a much lower bar and a much more achievable one.

Question 7

What is AI literacy for a business professional?

Accepted Answer

The practical ability to frame a request well, judge whether the answer is any good, know when something needs verifying, protect information that should not be shared, and recognize where human judgment is still required. It does not require coding, data science, or any understanding of how the models are built.

There is a great deal of confusion about what AI literacy actually means, and most people make it far larger than it needs to be. That does real damage. When executives believe literacy means technical mastery, two things happen. Some conclude they can never get there and delegate the whole subject. Others overcorrect and commit to initiatives that are poorly suited to their business, because they cannot yet tell a real opportunity from a compelling demonstration.

Practical literacy is a short list. Can you give the tool a clear task, a role, and enough context to work with? Can you look at what comes back and tell whether it is actually good, or only fluent? Do you know which claims need checking before you act on them? Do you know what information should never go into a tool that has not been approved? And do you know which parts of the work still require a person?

Executives need one layer beyond that, which is the ability to interpret what AI means for the business and to evaluate recommendations other people bring them. That is judgment applied to a new subject, not a new technical skill.

Success begins with literacy, not delegation. Someone who is AI literate can use the technology without treating it as either magic or as truth.

Question 8

Is it too late for my company to start with AI?

Accepted Answer

No, but waiting has a cost. Companies that delay end up playing catch-up, which is a difficult position for any executive. The window is still open. What matters more than speed is sequence: a company that starts deliberately in month twelve usually passes one that started randomly in month one.

The media is pushing hype and a lot of talk about a bubble, and business owners are reasonably asking whether the moment has passed. It has not. But timing is a choice with consequences. Not everyone will make the choice, and some will wait long enough that they have to run a catch-up game, which is always a hard place for a CEO to stand.

The more useful question is not whether you are late. It is whether the step you take next is the right one. Published estimates on AI project failure rates vary widely, but every serious source points in the same direction: a large share of initiatives stall or quietly disappear without producing measurable value. That makes the quality of your next decision far more important than its speed.

There is nothing wrong with action. The moment demands it. But action for its own sake is a non-starter. Taking the right path to your AI strategy will significantly improve your odds of capturing an advantage in your industry or service geography, and it costs less than recovering from a failed implementation.

Question 9

I know AI matters, but I am frozen and do not know where to begin. What is the first step?

Accepted Answer

Pick one real problem and use AI on it. You do not need a roadmap, a strategy deck, or confidence. The courage to begin does not come from having all the answers. It comes from having work that matters more than the discomfort of starting.

My first computer sat in the box for 42 days. I bought it in 1983 for about $10,000, most of it financed, and then left it unopened in the corner of my office. I had no technical background. I understood systems, but staring at that box I was completely frozen.

I finally opened it, not because I suddenly felt ready and not because I had mastered the manuals, but because I had work that mattered and staying frozen was no longer an option. I picked one problem and used the machine to solve it. That one problem became the path forward. The tool became familiar. Everything after that came easier.

I watch senior executives face the same moment with AI today. The overwhelm is real. The diversity of tools, the speed of change, the sweeping generalities coming from every direction: it adds up to genuine systems shock. It is not a character flaw. It is the predictable result of a market that is much better at generating noise than at helping anyone find an entry point.

There is a fair objection here, since a great deal of this library argues for diagnosing before you commit. Both are true and they operate at different scales. An organization about to spend real money across departments needs to understand its own operation first. One executive who has never used these tools needs to open the box. Personal experimentation is how you build the judgment that makes you useful in the larger conversation, and it costs almost nothing to be wrong.

If you are staring at your own boxed computer moment, you do not need to master the cockpit. You need to solve the next problem in front of you. Not with a pilot program. With a problem.

Question 10

Is AI realistic for a small or mid-sized company, or only for large enterprises?

Accepted Answer

It is realistic, and in some ways easier. Smaller companies have shorter decision paths and less legacy complexity. The constraint is not budget. It is knowing where to look. The screening test is the same at any size: can it be done cost effectively, is it commercial grade, and does it add real value?

AI is not just for technology giants. But the way it is being marketed makes it feel that way, because most of the coverage and most of the case studies come from companies with resources no mid-market business has.

The three-part screen I apply to any proposed application does not change with company size. Can it be done cost effectively at your scale? Is the underlying tool commercial grade, meaning will it hold up against the volume and variety of real customer interaction rather than a controlled demonstration? Does it add value that you can name and measure?

Plenty of impressive applications fail the first or second test for a company with thirty employees, and there is no shame in ruling them out. Plenty of unglamorous ones pass all three and produce returns a larger company would not bother to chase.

It is also worth saying plainly that most organizations do not need a large build at all in their first year. What they usually need is training that gives the leadership team a shared vocabulary, an audit that shows where the money is leaking, and a policy conversation that surfaces exposure nobody had considered. Those cost a fraction of a platform commitment, and they make every dollar spent after them smarter. Nobody needs another presentation explaining what AI is. They need help finding where to begin and where the return actually lives.

The advantage a smaller company has is speed of decision and proximity to the work. The owner usually knows exactly where the pain is. What is missing is a structured way to test whether AI is the right answer to it.

Question 11

Should we wait until AI matures before we commit?

Accepted Answer

There is never a perfect time, and waiting for certainty usually means waiting past the opportunity. There is never enough data to be certain, and by the time there is, competitors have generally taken the high ground. The advantage goes to leaders willing to act on sufficient rather than complete information.

In 1983 I bought my first computer system. It cost about $10,000, and $8,000 of that came from a bank loan.

The second sentence is the part worth pausing on. In 1983, an individual walking into a bank and asking to borrow money for computer equipment was essentially unheard of. There was no established category for the request, nothing comparable on file, and no way for the bank to know what the equipment would be worth in three years. I got the loan because one banker was willing to make a judgment ahead of his own institution's evidence. That was rare then, and that kind of banker is rare now.

That loan is a large part of why I have the career I have. Not because the machine itself was remarkable, but because owning one early meant I spent the following years learning something most of my peers would not touch for another decade. The educational leap compounded. The software business, the systems work, and eventually this work with AI all trace back to a decision someone made on incomplete information.

Which is the honest answer to the question. Waiting for certainty usually means waiting past the opportunity, because there is never enough data to be certain and by the time there is, someone else has taken the ground. In the early days of personal computing the debate was about the killer app, and it turned out to be spreadsheets and word processing. With AI we are still looking for that, and nobody honestly knows what it will be.

What I do know is that uncertainty is the constant and clarity plus execution is the advantage. Many executives freeze. Some chase every new thing. A few focus, move deliberately, and win. You know more about your customers and your business than any vendor does, and applied deliberately that is worth more than waiting for the market to settle.

It is worth noticing that the position you are in right now, deciding about AI without enough information, is the same position that banker was in. He decided anyway, and he was right.