The AI Readiness Toolkit
Three things you can run on your own operation.
A readiness self-assessment, a shortlist of where AI tends to pay off first, and a one-page governance starter. Use them with or without us. Nothing here needs a vendor to be useful.
01 / The AI Readiness Toolkit
Readiness self-assessment
Sixteen questions across the four foundations that decide whether AI scales or stalls. Answer honestly. It takes about ten minutes and the only person it can mislead is you.
Score each question
- 0
- No. Not true today, or nobody knows.
- 1
- Partly. True in places, not consistently.
- 2
- Yes. True, and you could show someone.
Architecture
Whether your systems can actually connect to anything new.
- 01.1Can your ERP or business system be reached programmatically, through an API or a supported integration, rather than only through screens and exports?
- 01.2Do you know who inside or outside the company is able to build an integration against your core systems?
- 01.3When a new tool has been added in the last three years, did it connect to existing systems or did it become another place to check?
- 01.4Is there a test or staging environment where a change can be tried before it touches live operations?
Data
Whether the information a system would need is clean and reachable.
- 02.1Is there one agreed place where current pricing lives, or does it depend on who you ask?
- 02.2Can you pull the last twelve months of quotes or orders into a single file without manual reconciliation?
- 02.3Are your part numbers, customer records, and product data consistent enough that the same item is not recorded three different ways?
- 02.4Do the exceptions, the rules that override the standard price or process, exist anywhere other than in one person’s head?
People
Whether the team is set up to use it once it exists.
- 03.1Is there a named person whose job includes making a new system get used, not just get installed?
- 03.2Have your frontline people been told plainly whether AI is meant to assist them or replace them?
- 03.3Is there at least a small group who would volunteer to go first rather than having to be pushed?
- 03.4When a process changes, does anyone check months later whether the new way actually stuck?
Governance
Whether someone owns the decisions and the quality.
- 04.1Is there a written answer to what staff may and may not put into a public AI tool?
- 04.2Does one named person or group decide which AI tools get adopted?
- 04.3Before an AI initiative starts, is a specific business metric named that it is supposed to move?
- 04.4Is there a defined point where AI output must be reviewed by a person before it reaches a customer?
Reading your score
Total of 320 to 12
Foundation first
A pilot started here will stall, and the stall will be blamed on AI rather than on the missing foundation. Fix the weakest pillar before you buy anything.
13 to 21
Ready for one workflow
You have enough to make a single high-volume workflow work end to end. Do not spread wider until that one is running without you watching it.
22 to 32
Ready for core systems
The foundations are there. The constraint is now sequencing and governance, not capability. Move past pilots and into the systems the operation runs on.
Read the pillar scores as well as the total. Any single pillar at 3 or below out of 8 is your binding constraint, whatever the total says. The four foundations do not average out. The weakest one sets the ceiling.
02 / The AI Readiness Toolkit
Where AI pays off first
The places AI tends to pay off first in an electronics manufacturing or distribution operation, in the order we would sequence them, with what has to be true before each one works.
Quote and proposal generation
It is high volume, highly repetitive, and already slow. In most electronics distribution and manufacturing operations a multi-line quote is assembled by hand from an ERP export, a price sheet, and the memory of whoever knows the exceptions. The work follows a pattern, which is exactly what AI handles well, and the cost of the delay is measurable in deals you can name.
What has to be true first
- Current pricing is reachable from one place.
- The last twelve months of quotes can be exported.
- A person approves every quote before it leaves.
How you know it worked
Quote turnaround time, and quotes sent per rep per week.
Technical answers and product knowledge
Field and inside sales people lose hours waiting on answers that already exist somewhere: a datasheet, a past ticket, an application engineer who answered the same question last quarter. Retrieval over documents you already own is one of the most reliable things this technology does, and the source can be shown next to the answer so a rep can check it.
What has to be true first
- Documentation exists somewhere reachable, even if it is messy.
- Someone can decide which sources are authoritative.
- Answers show their source so a rep can verify before repeating it.
How you know it worked
Time from question asked to answer in hand, and repeat questions to experts.
Channel and customer self-service
Distributors, dealers, and reps ask the same questions about stock, pricing, lead time, and order status, and every one of those questions consumes an internal person. This one pays off later than the first two because it is customer facing, which raises the quality bar. Do it third, not first.
What has to be true first
- Stock and order status are reachable in near real time.
- You are willing to publish pricing rules to the channel.
- A clear escalation path to a human exists and is fast.
How you know it worked
Inbound status and pricing enquiries handled without an internal person.
And the one that comes later
Executive revenue visibility, one consolidated picture across CRM, ERP, and finance, is usually the fourth. It is valuable, but it depends on the systems underneath it being trustworthy first. Build it after at least one of the three above is running, not before.
Where not to start
- Anything where the volume is low. If it happens twice a month, automating it saves nothing and teaches you nothing.
- Anything with no owner. A workflow nobody is accountable for will not survive first contact with a busy quarter.
- Anything where the rules were never written down and the person who knows them is about to retire. Capture the rules first, then automate.
- Anything customer facing on day one. Earn the quality track record internally before you point it outward.
03 / The AI Readiness Toolkit
One-page governance starter
Seven decisions. Fill in the blanks and you have an AI policy that fits on one sheet, which is the only kind anyone reads. Written to be completed in a meeting, not a quarter.
Who decides
Without a named owner, AI adoption happens in pockets and no one can stop a bad one or scale a good one.
AI tool decisions are owned by ____________________, who consults ____________________ before anything touches a core system.
What is allowed
Staff will use AI whether or not you have a policy. The only question is whether they are guessing.
Approved tools for general work are ____________________. Anything not on this list needs approval from the owner above.
What never goes in
This is the single highest-consequence line in the document, and the easiest one to write.
The following never goes into a public AI tool: customer pricing, supplier agreements, personal data, unreleased product information, and ____________________.
What a person must review
The review point is what separates a useful draft from an incident. Name it before you need it.
A named person reviews AI output before it reaches a customer in these cases: ____________________. Their approval is recorded.
How quality is checked
Quality that is nobody’s scheduled job silently degrades until someone complains.
Every ____________ weeks, ____________________ samples ______ outputs and records what was wrong. Findings go to the owner above.
What number it moves
An initiative with no named metric cannot be judged, so it never gets stopped and never gets scaled.
This initiative exists to move ____________________ from ____________ to ____________ by ____________.
When you stop
Deciding the kill criteria while you are still calm is the cheapest decision on this page.
If ____________________ has not happened by ____________, we stop and reassess rather than extend.
If you only do three
If you fill in nothing else on this page, fill in decisions 01, 03, and 06. An owner, a hard line on data, and a number. Those three prevent most of what goes wrong, and they fit on a single sheet you can put in front of a board.
If the assessment turned up more questions than answers, that is the useful outcome.
The Revenue Impact Diagnostic is 2 to 3 days on-site inside your business doing this properly, against your real systems and data rather than a self-scored sheet. $45,000. You leave with a priced map of where AI pays off in your operation, whether or not you build it with us.
USL Systems. uslsystems.co