AI consulting for business leaders. Right-size your automation.
Everyone is shouting. Nobody says where to start.
You have heard that most AI pilots fail and that the prices keep rising. You
have also heard that AI will save us all, and that it will end us. Holding
back is a reasonable response.
The answer is in the middle. AI is becoming a utility, like electricity and
running water. Start with the work you already do that costs you time and money.
AI Triage is a two-minute check that sorts your business
processes into the ones AI should take on first and the ones to leave alone for now.
All clicks, no email needed.
The wrong engine for the job. Right-sizing means AI for the judgment,
traditional tools for the heavy lifting, and the power matched to the work.
AI is a hammer. Not everything is a nail.
Since ChatGPT arrived in 2023, what people can get AI to do has grown every
day. It is changing the world the way electricity, steam, computing and the
Internet did. There is no going back.
But AI added to a broken process leaves it broken. AI asked to do what it is
bad at is wasteful, expensive and prone to error.
Step back. Look past the noise at the processes your business actually runs.
If a process is not documented, and people cannot do it today, it is not
ready to automate. Get it working for people first. Then hand it off, and
free them for the work people do best.
Your people want to work with AI, and many already are. It shows up in
four ways, and all four are reasonable.
Remarkable gains
Some people have rebuilt how they work and are far more productive than a year ago. Most of what they know lives in their own habits, and nobody else gets the benefit.
Impressive toys
Clever things built in an afternoon that nobody but the person who wrote the prompt will ever use. Good practice for them. Not yet a change to the business.
More than a text generator can give
Expecting judgment, exact arithmetic or perfect recall from something that generates plausible text. The disappointment is predictable, and so is the fix.
A better search box
Prompting reluctantly, and using AI as a search engine or an answer engine. It is very good at both. But is that what the business needs from its people using AI?
Individual enthusiasm does not add up to organizational change on its own.
Somebody has to decide what the business needs from AI, own the result, and
use it in front of everyone else.
That is a top-down job, and we work with leaders who have personal skin in the game.
The whole leadership team, on one call
A strategy for AI is three decisions, and in most companies three different
business leaders hold them. The strategic one is what AI is for. The tactical one
is what it is wired to. The operational one is who checks the result.
Typically that is the CEO, the CTO, and the COO, and they have rarely put
their answers side by side.
If a person cannot do the process end to end, even slowly, a machine will
not do it faster.
Business schools have taught it for a century: make the process work for
people first, then speed it up. Clear steps, clear ownership, a clear
definition of done. Only then is it worth handing to a machine, and by then
you know exactly what you are asking it to do.
So we start with the business problem, not the technology. What is the job,
who does it today, and where does it actually get stuck?
Right-size the engine to the job
A language model produces text one small piece at a time, and that is
remarkably powerful. It is also only one of several proven ways to automate
business work. Decades of older tools came before it: systems that search a
company’s documents, organize what a business knows, and apply fixed
rules. They give the same answer every time, you can check their work, and
they are cheap to run. The skill is knowing which to use where.
Step 1
Let AI do the language work
A language model produces text, and for reading, drafting, summarizing, sorting and deciding what to do next, it is exactly the right tool.
Step 2
Let proven tools do the computing
Adding a column of numbers is arithmetic, not prose. A spreadsheet or database does the math exactly, every time. A model can return a plausible-looking answer without doing the sum. We have the AI direct the real tool.
Step 3
Retrieve, then read
Search systems, and the ways businesses organize what they know, have spent decades getting the right record in front of the right reader. Use them to find it, and use AI to read it.
Step 4
Start small and adjust
Baseline the process, automate a slice, and check the cost and the errors against it. Keep what works. Mistakes stay small, early and cheap.
Some work needs an AI. Some needs a spreadsheet. Most needs both, with
something sensible in charge.
What makes it hold up
The demo is the easy part. Anything that runs your business has to
survive contact with your data, your people, and your auditor.
It knows your business
Generic AI gives generic answers. We ground it in your own records so what comes back is about your company, not about companies in general.
It shows its work
Every answer traces back to the record it came from. Your people check it in seconds instead of taking it on faith, which is what earns trust.
It stays inside the lines
Who can see what, what it is allowed to do, and a log of everything it did. The controls your auditor will ask about, built in from the start.
It gets tested
Before it runs and after, because the models change underneath you. Last quarter’s result is not a warranty on next quarter’s.
How we work
Start small, prove it works, then decide how far to take it. The approach
is the same for a bank, an agency, a drugmaker or a factory.
Applied Relevance also makes a product. Epinomy WMS runs a working warehouse:
it drives Universal Robots arms, talks to PLCs over MODBUS TCP, and deploys
air-gapped for sites that will never put their operations on someone else’s cloud.
We mention it because it keeps us practical. Software that has to run a
warehouse teaches you quickly which jobs need AI and which need a plain,
reliable program.
Pepsi, Coca-Cola, Quaker, Kodak, The Home Depot, Lowe’s
Industry and manufacturing
Siemens, Alcoa, Mercedes-Benz
Professional services
Deloitte, KPMG, PricewaterhouseCoopers, Ernst & Young, McKinsey & Company, Booz Allen Hamilton
Built by practitioners
We have spent thirty years in the technologies that came before: enterprise
search, organizing what businesses know, and managing and retrieving
information at global scale, across public sector, financial services,
pharmaceuticals, telecommunications and more. The tried-and-true techniques
still work, and they pair well with what AI does best.
We work in whatever you already run. We get things wrong too, so we build
in stages small enough that mistakes show up early and cost little. Part of
our value is knowing what not to build yet, or what to wait six months for
because it is about to arrive inside software you already pay for.
We take on a small number of clients at a time.
The first step is AI Triage: a two-minute check of which of your processes AI
should take on first. If you want to go through the result with someone, it
offers a 40-minute video call. No slides, no pitch.
