AI & automation
Practical AI, including the part where the answer is no.
Some problems are genuinely an AI problem. More of them are a redesigned process, software you already own, or work that should stop entirely. Knowing which is the entire job.
How should a business use AI?
A business should use AI where the work involves language, judgment at volume, or pattern recognition that people do slowly and inconsistently — and should not use it where a simpler rule, an existing system or removing the work would do. CompanyWRX inventories what is already in use, separates the genuine cases from the rest, and puts ownership, security and continuity around whatever stays.
What the work involves
- Inventory what is already in use — including the tools people built quietly and never mentioned.
- Separate the problems where AI is genuinely the answer from those where it is a process change, existing software, or stopping the work.
- Build the small number of things that pass that test.
- Put governance around it: who owns it, where the data goes, what happens if that person leaves, and what it must never be allowed to decide alone.
- Set the accuracy bar before building, and measure against it afterward.
What you hold afterward
- An honest map of where AI helps in your business and where it does not.
- Working automation on the cases that justified it.
- Governance that lets people keep building without the company depending on a personal account.
Encourage builders. Govern what they build.
When this is the wrong thing to buy
If the goal is to be able to say the company uses AI, do not spend the money. And if the data it would need is scattered, contradictory or wrong, fix that first — AI applied to bad data produces confident bad answers, which is worse than no answer.
Questions
AI automation, in practice.
Is our data safe?
That depends entirely on the architecture, and it is a decision rather than a property. What data leaves your environment, to which provider, under what retention terms, is designed deliberately and written down. If nobody can answer those three questions about a tool already in use, that is the first finding.
What if the AI gets it wrong?
It will, so the design question is what it is allowed to do unsupervised. Anything with a real consequence gets a person in the loop. Anything reversible and low stakes may not need one. Deciding that explicitly is the work.
Our team already uses AI tools we did not approve. Is that a problem?
It is a signal — those people found real friction and solved it. The risk is not that they built something; it is that the company now depends on a personal account nobody else can access. Govern it rather than banning it.
Where this usually leads
This rarely arrives on its own.
Somebody spends every Monday morning turning one report into another report.
If the outcome is what you are after
- Automate a businessSomebody spends every Monday turning one report into another report. Somebody else re-types the same job details into a second system. Neither of them can stop.
- Modernize a businessThe business runs well and looks like it did in 2014. Customers ask questions the website should have answered, and your team keeps a spreadsheet because the system cannot do it.
Usually done alongside
- Business automationAutomating the repetitive work that runs a company — follow-up, scheduling, reporting, data entry and handoffs — after the process is worth keeping.
- Operational automationAutomating the operational gaps that quietly cost money — unreturned calls, estimates nobody followed up, jobs never invoiced, leads that went nowhere.
- Custom softwareInternal applications, customer and employee portals, dashboards and mobile apps — built only when buying, configuring or integrating genuinely will not do the job.
Software that sometimes carries this
Not sure this is your problem?That is usually the most useful thing to find out first.
Answer thirteen questions about how your company actually runs and get a reading on each area — brand, website, search, reviews, software, automation, payments, operations, people, data and continuity.
Typically owner-led companies from first idea to around 500 people.