AI Guides › Playbooks
By Nigel Guy · 7 min read
Most people judge their AI tools on one question: is this the best one right now? That question gets re-answered every few weeks by a new benchmark, so you either keep switching or stop asking at all and drift on habit. Neither tells you which subscriptions are worth building your work around and which are already quietly fading.
The rule: keep a tool because it is better at a specific job you can name, not because it was first or familiar, and know in advance what would make you leave it.
Below: the Durability Ledger, then our stated read on seven common kinds of tool. That read is a bet, not a benchmark.
Products in this space do not just get worse; some stop existing. When Humane sold its assets to HP in February 2025, its Ai Pin stopped connecting to Humane's servers on 28 February 2025 and, according to reporting at the time, customer data on its cloud was deleted. Less dramatically, the models underneath tools you already use get retired on a schedule. Anthropic's model deprecations page lists models moving from Active to Deprecated to Retired, with at least 60 days' notice for publicly released models, after which "requests to retired models will fail". Other model providers retire models too; check their own deprecation pages.
So durability has two halves: will the company and product survive, and will your work survive if it does not?
One row per tool you use or pay for. Five columns.
| Column | What you write | Why it matters |
|---|---|---|
| Job | The specific task it does for you, in one line ("drafts client proposals from call notes"), not the category ("writing") | A vague job makes every rival look like a replacement |
| Why it wins | One letter: B (genuinely Better at this job), F (First and familiar), N (Nearest, because it is bundled into something you already pay for) | F and N tools are the ones that fade without you noticing |
| Exit cost | Can you export your data? Are your prompts, files and settings stored somewhere you control? Low / Medium / High | High exit cost on an F tool is the worst combination |
| Replacement trigger | The concrete thing that would make you switch ("a rival does X at the same quality inside the tool I already use") | Decided now, while you are calm, not during a launch-week rush |
| Review date | When you will look at this row again (six or twelve months) | Stops the ledger becoming a one-off exercise |
These are categories, not brands, because a five-year bet on a specific product name is guesswork. The verdicts are our opinion at time of writing.
| Kind of tool | Our read | Why |
|---|---|---|
| General assistant (ChatGPT, Claude, Gemini and similar) | Likely to survive as a category; the specific leader may change | Big labs with their own models. Your risk is lock-in to one, not its disappearance |
| Coding agent | Likely to survive, likely to consolidate | Tied closely to the model makers; small wrappers are exposed |
| Thin writing or "AI copywriter" wrapper | Already fading for most people | Mostly F: a prompt and an interface over someone else's model, now matched inside general assistants |
| Image and video generation | Survives, churns fast | Real quality differences, but the leader has changed repeatedly |
| Meeting notetaker | At risk of absorption | Increasingly built into the video-call and office suites you already pay for, which makes standalone tools an N-loser |
| AI search | Uncertain | Genuine B for some research jobs, but competing directly with search engines and general assistants adding the same feature |
| No-code automation (Zapier, Make, n8n and similar) | Likely to survive | Value sits in the integrations and your built workflows, not the AI layer, though your workflows are the exit cost |
Paste your list into an assistant with this prompt. Fill in your tools, what you use each for, and roughly what each costs you.
You are a sceptical technology adviser helping a small UK business review its AI subscriptions. Your job is to test my choices, not to reassure me.
My tools, one per line, as "tool name: what I use it for: monthly cost in £":
[TOOL_LIST]
My business and main type of work: [BUSINESS_CONTEXT]
Do this in order:
1. If any tool is missing its job or cost, ask me for it before continuing. Do not guess.
2. For each tool, classify why I probably keep it: B (better at a specific job, say which), F (first and familiar), or N (nearest, bundled with something I already pay for). Explain in one sentence.
3. Flag any two tools doing the same job.
4. For each tool, state what a competitor would need to offer for me to switch within five years, as a concrete, checkable condition.
5. Note anything you are unsure about, such as current features or pricing, and tell me to verify it rather than stating it as fact.
Output a table with columns: Tool | Job | B/F/N | Overlap | Replacement trigger | Confidence (high/medium/low). Follow it with no more than three plain sentences on what I should cut or review first.
Before answering, check: have you defended any tool simply because I listed it? If so, revise.
A two-person bookkeeping practice, invented for illustration, lists five tools: a general assistant, a standalone AI notetaker, an AI copywriting app, an automation platform and an image generator.
Two subscriptions go, one tool gets an exit plan, and the review date is set for twelve months.