What is AI Slop? A Business Buyer's Checklist for AI Tools and Agencies

AI slop is a bubble symptom: capital and attention reward the label AI, so the label gets applied to products that would never survive on merit alone. "A lot of companies raise money because it says AI. A lot of companies try tools because it says AI," says Hezarkhani (Arman Hezarkhani, co-founder of AI transformation firm Tenex, 2026). He opens his conference talks with two questions. First, raise your hand if you think AI is the most transformative technology ever built, and most hands go up. Then, raise your hand if you think we are in a bubble, and most hands go up again. His verdict: "Good job. You're both right." Both things can be true at once, just as they were for the internet around 2000. A transformative technology and a speculative bubble can coexist, and the bubble is what produces slop: tools built to raise money or ride a trend rather than solve a problem. The clearest symptom is the abandonment rate. Gartner predicted at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025 (Gartner, 2024). S&P Global found the share of companies scrapping most of their AI initiatives jumped to 42%, up from 17% a year earlier (S&P Global Market Intelligence, 2025). Buyers are not foolish. They are being sold to faster than they can evaluate.

Inside a business, AI slop looks like licences nobody uses, pilots that never reach production, and a retainer that outlives the results. The pattern repeats across industries. A tool is bought after a polished demo. The team tries it for a fortnight. Usage collapses because the tool solved nobody's actual problem. The licence renews anyway because cancelling means admitting the mistake. More than 80% of AI projects fail, roughly twice the failure rate of non-AI technology projects (RAND Corporation, 2024). "AI is one of the most transformational technologies that has ever existed, but it is incredibly difficult to adopt as a business. It is actually way easier to adopt as an individual," says Hezarkhani (Arman Hezarkhani, co-founder of AI transformation firm Tenex, 2026). One person can swap a search engine for a chatbot in an afternoon. A business has to change workflows, permissions, data handling and habits across dozens of people. Slop is what you get when a vendor sells you the afternoon version and leaves you to discover the organisational version alone. The tell is the measurement vacuum. Nobody recorded how many hours the task took before the pilot, so nobody can prove anything afterwards. The pilot cannot fail, but it cannot succeed either. It drifts into the pilot graveyard: not cancelled, not shipped, just quietly renewed.

Slop sells the label and skips the baseline. Real value starts from a measured problem and ends with you owning the result. Put any offer in front of these two lists. Signs of slop: - The pitch leads with the model name, not your problem - No baseline was measured before the pilot started - The demo works on sample data but never touches yours - Pricing scales with seats, not with outcomes - The vendor cannot say which parts are AI and which are ordinary code - Nobody on your team could run it if the vendor disappeared tomorrow - Success is defined as adoption or engagement, never hours or dollars Signs of real value: - Starts from a named, quantified problem measured in hours, dollars or error rates - A working trial on your real data before any contract is signed - A plain answer to which steps use AI and which use boring, reliable code - Fixed scope, fixed price and a production handover date - Your own staff are trained to run, fix and extend the system - Success metrics agreed in writing before work begins - The vendor is happy for you to own everything when they leave The pattern under every item on the left is the same: the seller keeps the knowledge and you keep the invoice. Every item on the right moves knowledge into your team. That single asymmetry predicts the widely cited 95% pilot failure figure (MIT, 2025) better than any feature list.

Before paying any AI vendor or agency, ask ten questions. Slop sellers rarely survive the third one. - 1. What problem does this solve, in one sentence, without using the word AI? - 2. Would this still be worth buying if the label said software instead of AI? - 3. Which parts are actually AI, and which parts are ordinary code? - 4. Can we trial it on our real data, in our real workflow, before signing anything? - 5. What baseline will this improve, and who measured that baseline? - 6. What does success look like in hours or dollars (AUD), and on what date will we measure it? - 7. Can you name two customers our size who have run this in production for at least six months? - 8. Who owns the system, the data and the intellectual property when the engagement ends? - 9. What breaks first, and who fixes it when it breaks at 2am? - 10. What is the total three-year cost, including licences, retainers and our own staff time? Question two is Hezarkhani's own filter: "If you would be willing to invest in this thing if it didn't say AI, will you do it? That is the main question. Just try to remove AI from it" (Arman Hezarkhani, co-founder of AI transformation firm Tenex, 2026). Only 26% of companies report moving past proofs of concept to generate tangible value from AI (BCG, 2024). The companies in that 26% are the ones who asked uncomfortable questions before signing, not after.

An agency is selling slop when the strategy deck is the product, the delivery team is junior, and everything you paid for stops working the day you stop paying. First, ask what you are actually buying. "A lot of those AI solutions are a 10-step process where nine of the 10 steps are just traditional code, and maybe one of the steps is AI," says Alex Lieberman (Alex Lieberman, co-founder of Tenex, 2026). An honest agency tells you this up front and prices accordingly. A slop agency charges AI prices for the nine boring steps and hides them behind the buzzword. Second, watch what they prescribe. "People feel like AI is a hammer and every problem is a nail, but that is just not the case" (Alex Lieberman, co-founder of Tenex, 2026). An agency that proposes an AI solution before it has heard your problems has already failed the checklist. The remaining flags: no fixed scope, no production handover date, no named senior engineer, case studies that are demos rather than running systems, and a model where every question routes back through them. If the plan requires you to keep paying the agency to operate what it built, you did not buy a system. You rented one. The stakes are real. With 95% of AI pilots delivering no measurable return, a figure widely cited since its release (MIT, 2025), the default outcome of a badly chosen engagement is nothing at all.

The alternative to buying slop is not avoiding AI. It is choosing a path where the system, or the capability to build systems, ends up owned by you. Path one: you have a specific system to build and you want it built properly, once. That is what 13Labs does through buildAgency: a fixed-scope production AI build by a senior engineer, with a defined price, a production handover date and your team owning the result. No retainer dependency, no black box. Details at 13labs.au/buildAgency. Path two: you want the capability inside your business permanently. Through buildAutomation, 13Labs trains two or three of your own staff, over six to twelve weeks, to diagnose, build and own the automations your business runs on. When we leave, the capability stays. Details at 13labs.au/buildAutomation. Both paths start the same way this checklist does: with a named problem and a measured baseline, not with a tool. The 5% of pilots that do deliver returns (MIT, 2025) share exactly these traits: a real problem, a measured starting point, and internal ownership of the result. The questions in this guide are free. Any vendor or agency worth paying will welcome them. Any that deflects them has answered for you.

**Is AI slop the same as an AI scam?** Not usually. Most slop is legal and sincerely marketed. It is simply useless for your situation. The practical test is not intent but fit: if a tool cannot be tied to a named, measured problem in your business, it is slop for you even if it works somewhere else. **How much does AI slop cost an Australian business?** Licences, integration work and distracted staff time commonly run into the tens of thousands of dollars (AUD) per abandoned tool, before counting the opportunity cost of a year spent on the wrong problem. With 95% of AI pilots delivering no measurable return (MIT, 2025), the default spend is large and the default return is zero. **What is the fastest test for AI slop?** Remove the word AI from the pitch and ask whether you would still buy it. Arman Hezarkhani, co-founder of AI transformation firm Tenex, calls this the main question (2026). If the offer only sounds valuable because of the label, walk away. If it still sounds valuable as plain software, keep asking the checklist questions. **Should we buy AI tools or build our own capability?** Buy tools for commodity problems, but only after a working trial on your real data. Build, or have capability built into your team, when the workflow is core to your business. 13Labs covers both paths: fixed-scope production builds through buildAgency (13labs.au/buildAgency) and team training through buildAutomation (13labs.au/buildAutomation). **Does AI slop only mean bad software?** No. The term began as a label for low-quality, mass-produced AI content clogging search results and social feeds. In a buying context it has widened to cover any AI product or service sold on hype rather than results. Both senses share one root: volume without value.