The AI & business tools I'd actually pay for — organised by job-to-be-done, so you can build a lean stack in an afternoon instead of a month. Free PDF, straight to your inbox.
Drop your email — I'll send the full stack now, then one honest issue a month:
No spam. Unsubscribe anytime.
Curated from 190+ tools I've reviewed · by an ex-Deutsche Bank operator · honest, no fluff.
Curated by Daniel Haket — 20 years in international banking (ex-Deutsche Bank), now reviewing the software he runs his own business on. About →
Software categories are a vendor convention. “CRM” contains tools that share a label and almost nothing else, while the thing you actually need — keeping track of who owes you a reply — is solved by four tools in three different categories. Picking by job avoids the trap of comparing twelve products that all technically qualify and none of which match what you were trying to do. It also makes the gaps visible: a stack organised by job shows you which work is currently done by nobody.
The most common mistake in a new stack is buying for the company you plan to be. Analytics before you have traffic, a CRM before you have a pipeline, an automation platform before there is a process worth automating — each is defensible in isolation and collectively they produce a stack that costs money every month to describe work that is not happening yet. Buy the tool the week the task becomes annoying. Annoyance is a better signal than planning, because it is evidence.
Some spending is worth doing before it hurts. Anything touching money, because mistakes there are expensive and slow to unwind. Anything holding data you would be unable to reconstruct, where a free tier with export limits is a future problem you are choosing now. And anything customers touch directly, because that is where a rough edge costs you more than the licence. Everything else can wait for the annoyance test.
“Free plan available” covers several different offers. SupaEasy’s free tier runs only on a development store, so anything live starts at $49 and the AI generator the app is named for begins at $99. Easy Advanced Translations gives free AI for fifty products, after which you either pay or supply your own OpenAI key and cover the tokens yourself. Neither is misleading; both mean the free tier is for evaluating, not for running. Check what the limit is measured in — records, seats, volume — and how quickly you will reach it.
A stack built from per-seat tools grows with headcount whether or not usage grows with it, and licences outlive the people who had them. A stack built from metered tools tracks activity, which is fairer and much harder to budget: Ava AI sells $29, $99 and $499 tiers that are really allowances of call minutes, billing $0.07 to $0.10 for each minute beyond. Neither model is wrong. Knowing which one each line uses tells you what the stack costs at twice your current size, and that is the only forecast worth writing down.
Enterprise plans bought for one feature, annual commitments on tools you have used for under a month, and any platform whose value depends on consolidating systems you do not yet have. Rippling’s pitch — HR, IT and finance on one employee record — is genuinely strong, and it only pays off once several of your systems live there. Run one module of a platform priced for several and you pay platform rates for a point tool.
Every tool added is another login, another invoice, another integration to maintain and another thing to migrate later. The cost of a stack is not the sum of its prices; it is the sum of its prices plus the attention it takes to keep running. A lean stack you can explain in one breath beats a thorough one nobody has audited in a year — and when it does drift, the stack auditor and the cost audit are here to pull it back.
It is the stack we would build today, written the same way as the reviews: what each pick is good at, what it is bad at, and who should choose something else. Some links are affiliate links, which changes nothing about the picks — the same twelve would be here if none of them paid. How we review →
The jobs generalise better than the tools. Invoicing, keeping track of customers, publishing, supporting people who bought from you — almost every small business does those, and the shape of the stack is similar. The specific picks lean toward software and services businesses, so treat the categories as the durable part and the names as a starting shortlist.
Pricing and positioning move constantly, so every underlying review carries the date it was last verified. Check the live pricing page before you buy; the number you see there is the one that counts, and it is often the annual rate rather than the monthly one.
Then this is the wrong tool and the Stack X-Ray is the right one: it takes what you already pay and shows you the shape of it. Build from scratch with this; diagnose what exists with that.