Stop Re-Explaining Your Company to AI
Most businesses are using AI like a very smart intern with amnesia.
Every time they open ChatGPT or Claude, they type the same setup:
“We’re a Calgary company. We do this. Our customers are these people. Here is our tone. Here are our services. Don’t make things up.”
Then they wonder why the output is generic.
Claude Projects is one of the simplest ways to fix that.
It gives Claude a dedicated place to learn the context behind a piece of work: your documents, your instructions, and the conversations connected to that work. Claude calls these self-contained workspaces “Projects.”support.claude
That may sound small.
It is not.
The real problem is not prompting
The average business does not have a prompt problem.
It has a context problem.
AI cannot produce a useful proposal if it does not know:
- What you actually sell
- Which customers you want more of
- What makes you different
- What you are allowed to promise
- How your team currently does the work
- Which documents contain the truth
This is why “just use AI” is bad advice.
The useful question is: What does the AI need to know before it can help us do this job well?
Claude Projects is built for that question.
What is a Claude Project?
A Claude Project is a workspace for one job.
You add the documents Claude needs. You tell it how to work. Then you keep related conversations inside that same project.
For example, you might build a project called:
“Sales Proposals for Our Industrial Services Team.”
Inside it, you could add:
- Your service descriptions
- Past winning proposals
- Customer types you want to attract
- Pricing rules
- Common sales objections
- Brand voice guidelines
- Approved case studies
- Legal or compliance constraints
Then you tell Claude:
“Write like an experienced advisor. Be concise. Ask for missing details. Do not invent results, prices, timelines, or customer claims.”
Now your AI is not starting cold every time.
It has a brief. A source of truth. A job.
Here is the shift
Some companies will upload random documents, ask Claude to “make us more efficient,” and call it an AI strategy.
Others will choose one expensive, repetitive workflow, give AI the right context, and measure whether the work gets better.
The second group will pull ahead.
Not because they have better prompts.
Because they redesigned the work.
That distinction matters. Successful AI adoption is less about layering a chatbot on top of existing habits and more about rethinking the workflow around people, process, and technology.coalitioninc+1
What should go into a Project?
Start with a business outcome, not a tool.
Bad starting point:
“Let’s create a Claude Project.”
Better starting point:
“Our sales team takes too long to turn discovery calls into strong follow-ups and proposals.”
Then build the Project around that specific work.
| If you want to improve… | Create a Project for… | Add these materials |
|---|---|---|
| Sales follow-up | Discovery calls and proposals | Call notes, winning proposals, services, objections, pricing guidance |
| Marketing output | Content and campaign planning | Brand voice, customer questions, existing content, offers, campaign goals |
| Employee onboarding | Internal knowledge and training | SOPs, policies, role guides, FAQs, checklists |
| Client delivery | A specific account or program | Scope, meeting notes, research, deliverables, project plans |
| Leadership planning | A strategic decision | Financial assumptions, market research, customer feedback, leadership notes |
The Project is not the value.
The value is deciding what belongs in it, what does not, who can use it, and what process it improves.
That is where most businesses get stuck.
A Calgary example
Let’s say you run a professional-services firm in Calgary.
You have good people. Good work. Good clients.
But every proposal starts from scratch. The sales team hunts through old files. Someone rewrites a previous proposal. Someone else checks the language. Deadlines get tight. The final version is “fine,” but not always consistent.
A Claude Project can help centralize the usable context:
- What you sell
- Who you serve
- Where you create value
- What proof you can use
- How proposals should sound
- What must be reviewed by a human
Then the team can use it to turn a discovery-call transcript into:
- A summary of needs and risks
- A list of missing information
- A draft scope of work
- A tailored proposal outline
- A follow-up email
No, the AI should not send the proposal without review.
Yes, it can remove a large amount of the blank-page work.
That is a much better use of AI than asking it for “10 LinkedIn posts about leadership.”
The biggest mistake: the document dump
Do not throw 200 files into a Project and expect magic.
More information is not the same as better information.
If the material is old, contradictory, poorly named, or full of exceptions, Claude has a harder job. So does your team.
Before you upload anything, ask:
- Is this current?
- Is this approved?
- Is this something we would want an employee to rely on?
- Does it help AI make a better recommendation or create a better first draft?
- Is it safe and appropriate to use in this tool?
Good AI systems are curated.
They are not digital junk drawers.
What Claude Projects can do
Claude Projects are available to all Claude users, including free users, who can create up to five Projects.support.claude
Paid Claude plans offer expanded project knowledge capabilities, including retrieval-augmented generation (RAG), which helps Claude locate relevant information within larger sets of material.support.claude
Team and Enterprise users can also share Projects and set view or edit permissions, allowing organizations to create more consistent, controlled AI workspaces.support.claude
That makes Projects useful for work such as:
- Building a company-approved sales assistant
- Creating a brand-aware content workflow
- Turning internal documentation into a practical knowledge resource
- Preparing leaders for meetings and strategic decisions
- Standardizing client delivery across a team
Do not skip governance
If you are putting client data, internal documentation, contracts, employee records, or commercially sensitive material into an AI tool, governance is not a later problem.
It is the first problem.
Decide upfront:
- What information is allowed
- What information is prohibited
- Who can access each Project
- Who reviews work before it reaches a client or the public
- Who updates the Project when the business changes
This is not bureaucracy.
It is how you avoid building a fast system that creates slow, expensive problems later.
Start with one workflow
You do not need an “AI transformation” program to start.
You need one workflow with enough pain that fixing it would matter.
Pick one that is:
- Repetitive
- Time-consuming
- Document-heavy
- Important enough to measure
- Low enough risk to test properly
Then define success before you begin.
Maybe success means:
- Proposals go out two days faster
- Sales follow-up is completed the same day
- New employees ramp faster
- Content requires fewer revisions
- Client-facing work becomes more consistent
- Your team spends less time searching for information
If you cannot describe the outcome, you are not ready to automate the work.
The question worth asking
The question is not:
“Should we use Claude Projects?”
The better question is:
“Which part of our business would perform better if AI understood our context?”
That is the starting point for practical AI adoption.
And it is where a proper AI workflow audit becomes useful: identifying the work worth improving, mapping the information the AI needs, setting the rules, and designing a process your team will actually use.
My goal is to help Calgary businesses, break down their challenges and turn AI tools into working systems, not another experiment that people forget about after two weeks.
