There is a lot of bad AI sales language in the market right now.
The easiest businesses to sell it to are the ones that know they have operational friction but do not have the time to translate that friction into technical questions. That is why owners and managers end up looking at proposals that sound impressive but do not actually explain what will get better.
If you are reviewing an AI automation proposal, the first thing to check is simple:
Does it understand your workflow, or is it mostly selling software theatre?
The red flags show up early
A weak proposal usually gives itself away fast.
Watch for this:
- it talks about tools more than workflows
- it promises broad outcomes without naming one concrete process
- it avoids questions about your current software stack
- it has no pilot plan
- it has no exception handling plan
- it cannot say who owns the system after launch
- it implies staff replacement as the main value story
That is usually enough to stop the conversation.
What a real proposal should include
At minimum, a serious proposal should define:
- the exact workflow being improved
- the current pain and why it matters
- what will stay human and what will be automated
- how the existing tools are involved
- how exceptions are handled
- how success will be measured
If those things are not clear, the project is still too vague to buy.
Beware the fake confidence move
One common pattern is the consultant who sounds very certain about AI but stays fuzzy on operations.
They will say things like:
- “we can automate most of this”
- “we’ll plug into your systems”
- “this should save a lot of time”
- “your team will be much more efficient”
That all sounds fine until you ask:
- which workflow?
- whose time?
- what data source?
- what happens when the input is incomplete?
- what happens when the workflow hits an exception?
If the answers stay vague, the proposal is not mature.
The wrong first project
Another red flag is when the first proposed build is too broad.
If the plan tries to automate multiple departments, multiple systems, and multiple edge-case-heavy processes in the first round, that is not ambition. That is a poor scoping decision.
The first project should be narrow enough to prove value and controlled enough that failure does not disrupt the business.
The local-business filter
This matters even more for small and midsize businesses in Lethbridge and Southern Alberta.
Most local businesses do not need a giant AI roadmap. They need one workflow fixed properly. If the proposal sounds like it was written for a US tech company or enterprise transformation team, it probably was.
You want something that understands:
- busy owners
- overloaded managers
- older software
- real-world handoffs
- small teams with no room for science projects
That is the standard.
What a good proposal sounds like
A good proposal is usually more boring than a bad one.
It sounds like:
- “Here is the first workflow I would target.”
- “Here is why it is worth fixing.”
- “Here is what the system would do.”
- “Here is what stays manual.”
- “Here is how we would test it.”
- “Here is how we would know it worked.”
That is what you should trust.
Final take
If an AI proposal cannot clearly explain the workflow, the ownership, the exceptions, and the success measure, do not buy it.
The right project should feel more operational than inspirational. That is usually the clearest sign the person proposing it understands how real businesses actually run.