Find the outcome you need.
Start with a symptom, a platform or the thing you want working.
With JavaScript, search filters the whole collection in your browser. Without it, use the linked pages and section navigation below; query filters need JavaScript.
Showing entries 1121–1160 of 1179 in this first collection.
Buyer collection
For a head of sales whose CRM nobody trusts: list what is wrong before anything merges or closes
Duplicates, blank fields, dead deals and unused properties, listed with evidence and sampled by your own person before a single record changes.
Buyer collection
For a marketing manager: keep bot leads out of the count and prove every email tool honours an unsubscribe
Hold junk submissions for review instead of routing them, check that opt-outs reach every tool, and watch your sending sources each month.
Buyer collection
For a customer-success lead: see renewals at risk with evidence, and hand accounts over with a brief you can trust
A weekly list of renewing accounts with agreed risk signals, handover briefs where every statement links to a CRM record, and win-loss reasons from notes or interviews.
Buyer collection
For a marketer: find the pages losing traffic, link your pages, and hear when ad spend jumps
Search Console analysis with the queries behind each fall, internal links from a list you tick, briefs for a human writer and read-only alerts on ad spend.
Buyer collection
For a founder with a security questionnaire due: answer from sources and leave the rest blank
A late-stage customer questionnaire shrinks if every answer has a source, unsupported rows stay blank, and one named person signs the final.
Buyer collection
For a UK supplier asked for Cyber Essentials: find the gaps before you pay an assessor
The scheme has five controls and an assessor who decides. A gap list and drafted answers tied to evidence show the work before you pay.
Buyer collection
For an operations lead with contracts in folders and a first data request to expect
Get the dates in your contracts and the steps for a subject access request written down and tested before you need them.
Collection
Security and compliance paperwork in order: from what is true to evidence you can show
A six-step order for small companies facing customer security reviews, Cyber Essentials, audits and data requests, with the job for each step and what stays with you.
Inspectable example
Pass-mark arithmetic: what 38 out of 40 does and does not show about an AI-drafted answer set
A table and a short script that turn a test result such as 38 of 40 into the lowest true pass rate it supports, with 95 percent confidence.
Troubleshooting guide
Answer a security questionnaire only from a source, and leave the rest blank
A short method for a late-stage customer questionnaire: one master sheet, a source for every answer, blanks for the rest, and a named signer.
Troubleshooting guide
Cyber Essentials in plain words: the five controls, the two routes and who decides
What the UK scheme covers, how self-led and supported routes differ, what Plus adds, and why no one can promise you will pass.
Troubleshooting guide
Subject access requests: the clock, the reasonable search and what you must document
What the ICO guide says about the one-month deadline, extensions, a proportionate search, exemptions and refusals, and where a written process helps.
Troubleshooting guide
Read-only roles for reviewing a Microsoft 365 or Google tenant, and why connected apps matter
Which built-in roles and scopes let a reviewer read sign-in and audit logs without changing anything, which role can also read app consent grants, what native access reviews need, and why connected apps matter.
Buyer collection
For a store owner answering the same support emails: write down what you promise, test it, keep money with a person
Write one answer sheet from your own pages, test it on 40 questions, say when a system is answering, and keep refunds and photo claims with a person.
Buyer collection
For a store owner fixing a catalogue in bulk: check what AI wrote, fill what is missing, make every change undoable
Find the patterns bulk AI copy gets wrong, fill missing fields only from text you hold, and export before every change so you can restore it.
Buyer collection
For a store owner who sets prices and reorders stock: a floor under every price, and reorder points that ignore sold-out days
Work a floor from landed cost, get alerts before any automation, and count only the days a product could be bought when you plan reorders.
Collection
Store AI checks in order: answers, product data, prices, stock, disputes
A five-step order for checking what AI and bulk tools do in your online store, with the guide or job for each step.
Troubleshooting guide
Six ways bulk AI product copy goes wrong, and how to find each before you import
Wrong-product text, cut-off sentences, made-up specs, overwritten fields, broken characters and feed drift, with a check for each pattern.
Troubleshooting guide
Where is my order: eight delivery states, and the check that stops a stale reply
Eight delivery states for reply templates, the four that need a person, and a pre-send check that compares each value in a draft with the live order.
Troubleshooting guide
A price floor from landed cost: the arithmetic, and why alerts come before automation
How to work the lowest price you will accept from cost, fees, returns and margin, with a formula, and why to set alerts before any automatic matching.
Inspectable example
Synthetic price floors: ten invented products, two below their floor
Ten invented products with a landed cost, a price and a worked floor, showing a sale-priced item and a stale bundle below the line.
Troubleshooting guide
Stocky closed on 31 August 2026: what reorder planning needs from your own sales
What store owners say they lost when Stocky closed, and the three things a reorder sheet needs: a point per product, sold-out days removed and a check against a simple average.
Inspectable example
Synthetic reorder point: one invented product, 8 sold-out days, 29 units of difference
One invented product over 28 days shows how counting sold-out days as zero demand gives a reorder point of 73 instead of 102.
Buyer collection
For a support lead about to switch on an AI agent: test it before your customers do
Write down what the bot must refuse and hand over, run a written test three times, and check the label and the way to a person before go-live.
Buyer collection
For an owner answering the same questions all day: write the answers once, with their sources
For an owner using a help desk: find what customers ask that no page answers, write saved replies tied to your policy, and only then consider a bot or drafts.
Buyer collection
For a support lead with hundreds of tags and slow triage: clean the list, then test suggestions
Count tag use, agree a short list with definitions, have two staff tag the same tickets, and only then ask a tool to suggest tags.
Collection
From a help centre you trust to a support bot you can switch on: the path in order
Seven steps in order: find the gaps, fix the facts, test the bot, check the handover, try hostile messages, go live yourself and keep checking.
Inspectable example
A blank support bot test set: four kinds of case, three runs and the fields to fill in
A table you can copy: one row per case with its kind, expected result, source page, severe-error flag and three run results.
Troubleshooting guide
What AI support agents charge per resolution, and what to check about "resolved"
Vendor list prices read on 12 October 2026, the arithmetic for a small inbox and the one question to ask about what counts as resolved.
Troubleshooting guide
Test a support bot before you switch it on: four kinds of case, three runs, written pass marks
How to build a test set from your own pages, why each case runs more than once, and how to read a pass without over-reading it.
Troubleshooting guide
Which support tags to keep: count, merge, define and check that two people agree
A short routine for turning a tag list nobody trusts into one people use the same way, with the figures from Apollo's own attempt.
Troubleshooting guide
What a hostile message can make a support bot do: three cases and the permission behind each
A ticket, a form field and a chat message have each been shown to make a support bot or agent leak data; the pattern is untrusted text plus access plus a way out.
Troubleshooting guide
A subject access request in your support email: what it looks like and the one-month clock
What the UK regulator says about requests for a person's own data: no special wording, any channel, one month, and staff who can recognise one.
Inspectable example
Synthetic CI buying comparison: a fee difference is not a scope or savings verdict
Compare invented DIY and tool effort with the existing £149 one-job and £295 monthly offers, keeping hypothetical workload, customer work and untested prices visible.
Inspectable example
Synthetic CI monthly ledger: six red attempts are not six accepted fixes
A made-up failure ledger separates repeat attempts, one repaired cause, a flaky test, an account action and an unmerged proposal before discussing a monthly allowance.
Troubleshooting guide
Who owns a flaky CI test? Separate investigation, quarantine approval and the product fix
An intermittent pass does not establish repair. Assign the evidence, temporary risk decision and actual test or application correction before buying a repeatable-job fix.
Troubleshooting guide
What does a red CI build cost? Separate runner usage, active work and blocked time
Use job-level usage and actual investigation effort without multiplying a red badge by every engineer's salary or treating included minutes as cash savings.
Troubleshooting guide
Your CI maintainer is leaving: transfer responsibility, not just workflow notifications
Before a supplier ends or an engineer leaves, assign the open failures, alerts, merge authority and credential revocation without assuming a green last run proves continuity.
Inspectable example
Synthetic invoice reconciliation: identical net lines, two different tax totals
Three invented small-value lines show how line-level and invoice-level tax rounding can disagree even when net values and line counts match.
Inspectable example
Synthetic invoice retry matrix: source operation, attempt and target draft are different IDs
Concrete made-up references specify when a retry retains one target draft and when changed contents, a second organisation or multiple matches need a separate decision.