AI made insight cheap. Now what?

AI has not lowered the bar for insight. It has removed the excuse for not doing it.

AI’s profile has risen fast, from novelty to boardroom curiosity to a live conversation inside agribusiness in the space of a couple of years. AI is genuinely good at organising, accelerating and surfacing themes from information a business already holds or can gather quickly, doing all three faster and more cheaply than a research team ever could.

But it is not a replacement for the harder parts of the job. Deciding what actually matters, exercising judgement about what is true and why, doing the conversations, the testing, the data collection needed to validate what it surfaces, all of that still needs to be done well. Used well, AI is a way to do far more of the groundwork yourself, faster than before. Used carelessly, it is an easy way to produce something that looks like insight without having earned the name.

I think this is what most agribusinesses are underestimating right now. Not whether the tool works, because it clearly does, but that the same willingness to generate a plausible-looking answer to almost any question is what makes sloppy use so easy to mistake for progress. Most agribusinesses are already sitting on more information than they realise. Emails from customers, notes from field visits, supplier correspondence, industry reports, call logs, feedback forms. All of it is there, while almost none of it has ever been properly analysed, not because it lacks value but because turning it into insight has traditionally needed time, budget and dedicated research resource that most agribusinesses simply do not have. AI has changed that equation. You no longer need a research team or a six week project to start generating real insight from what you already hold.

New Zealand’s own numbers back this up. Only 36% of New Zealand SMEs are currently using AI in their business, a figure that falls to 30% among the smallest firms. Most agribusinesses have not started, which means most of the advantage described below is still sitting there for the taking.¹

But…

AI is only as useful as the question you bring to it. Fired off without a clear purpose, it will happily generate summaries, comparisons and ideas all day long, most of which nobody will ever act on.

The agribusinesses getting genuine value from AI-assisted insight are the ones spending a few minutes up front deciding what “value” looks like: who this is for, what decision it needs to inform and what a useful answer would look like. Do that first and everything below gets sharper. Skip it and you end up with an impressive pile of output and no clearer idea of what to do next.

With that in mind, here are three things worth doing tomorrow.

Every agribusiness has a backlog of unread evidence. AI is the first tool cheap enough to read it.

AI is well suited to unstructured data. Every agribusiness has a backlog of it, rep visit notes, customer emails, call summaries, complaint logs, supplier correspondence. It is rich material, but it is also messy, inconsistent and too time consuming to code and theme by hand, which is exactly why it has historically just been left alone.

This is precisely the kind of task AI handles well. It can read through large volumes of text and surface recurring themes, sentiment and patterns in a fraction of the time a person would need, turning a folder full of files into something you can act on.

Try this tomorrow. Pick one existing source, a file or folder containing customer feedback or conversation notes is a good place to start. Ask an AI tool to identify the recurring themes, complaints or shifts in sentiment. Most people are surprised by what has been sitting there unnoticed.

But… do not simply dump the documents in and see what comes out. Decide first what question you are trying to answer and who is going to act on the answer. “What are customers telling us about our new pricing model?” will get you something useful. “Tell me what is in these emails” will get you a well organised list of things nobody asked for.

Competitive scanning used to be a luxury only the big players could afford. That excuse is gone.

Formal competitive scanning and market intelligence has always been something bigger organisations do and smaller ones skip. Doing it properly has traditionally been slow and expensive, well beyond what most agribusiness marketing budgets can absorb. Most agribusinesses end up making strategic calls with only a partial view of what competitors, adjacent sectors and international leaders are doing.

AI closes a lot of that gap. It can synthesise publicly available information into a fast, structured view of how others are approaching a particular issue, removing the need for hours of manual searching and reading.

Try this tomorrow. Choose a specific issue that matters to your business right now, sustainability claims, pricing strategy or digital engagement are common ones. Ask AI to pull together a quick comparison of how three to five relevant players, local or global, are approaching it.

But… competitive scanning for its own sake is a distraction, not insight. Before you run the comparison, be clear on why you are doing it and what you would actually do differently depending on what you find. A well defined question like “how are leading dairy exporters positioning sustainability to European buyers?” will sharpen the output considerably more than “what are competitors doing”.

Every team has hunches built from years in the game. Almost none of them ever get tested.

Every agribusiness team carries instincts built from years in the game: younger customers seem less loyal, the new labelling seems to be confusing less frequent buyers, the best customers seem to be quietly looking at other options. These hunches are often right. They usually stay hunches anyway, because there has never been a formal process for checking them and no dedicated insight team to do it.

This is somewhere AI adds real value. It is good at helping you take a vague instinct and sharpen it into a specific, testable question, then pointing you toward what evidence would confirm or kill it. Instead of “I have a feeling about this,” you end up with something you can go and do, or a plan you can take to the business to resource.

Pick one hunch you hold about your customers, distributors or the market. Ask AI to help turn it into a precise question, along with a shortlist of what data, feedback or conversations would settle it one way or the other. That is worth doing before the week is out, not filed away for “someday.”

But… this only works if you are honest about what you would do with the answer. If you are not prepared to act differently depending on what the evidence shows, you are not testing a hunch. You are looking for confirmation. Be clear going in on what decision this is meant to inform. Be genuinely open to being surprised.

There is a pattern running through all three of these. AI dramatically lowers the cost and effort of a particular type of insight work, but it does not remove the need for discipline. If anything, it raises the importance of it. When insight was slow and expensive to produce, scarcity forced you to be selective about what you looked into. Now that it is faster and cheaper, the risk runs the other way, with output generated for its own sake, no clear owner, no clear audience, no clear purpose behind it.

So before you open a tool tomorrow, come back to the three questions this piece opened with.

1. Who is this for? Name the person who will read the answer and act on it, not a general audience. If you cannot name them, the exercise is not ready to start.
2. What decision will it inform? If there is no decision attached, you are producing content, not insight. Attach the decision before you attach the prompt.
3. What would a useful answer look like? Decide this before you start, not after you see what comes out. Write it down if that is what it takes to hold yourself to it.

Answer those first and each of the three approaches above will give you something sharper than a scattergun run at the tool ever could.

Picking one of these this week is a good first step. If you want a hand thinking through what “value” looks like for your business, a short insight audit is usually the fastest place to start. It is exactly the kind of work we do for agribusinesses at Tracta.