3 things worth configuring so your AI actually knows your business
Let’s say you open ChatGPT or Claude to draft a client proposal. The first thing you do is spend fifteen minutes pasting in your company description, your service offerings, the tone you prefer, the client's background, and the project scope.
By the time the AI has enough context to be useful, you've already done most of the heavy lifting yourself.
Forty-five minutes later, you have a decent first draft. But you spent more time loading context than you did thinking about the actual proposal.
Tomorrow, you'll do the same thing again. Paste, explain, re-explain, hope the output lands close enough to be useful. That cycle is the single biggest reason most consultants and founders still think AI is "interesting but not quite there yet."
The tool isn't broken. It just has no idea who you are.
What Has Changed So Far in 2026
ChatGPT, Claude, and Gemini now all offer some version of persistent memory. For most users, the feature showed up as a settings toggle or a notification they dismissed without a second thought.
Here's what that feature actually does. When configured, AI memory stores key context about you across sessions, including your role, your preferences, and your active projects. Instead of starting every conversation from zero, the tool pulls that context forward automatically.
Think of it like onboarding a new contractor versus working with someone who already knows your business. The contractor who knows your standards, your clients, and your constraints produces better work faster. AI memory works the same way, except the onboarding happens once and carries forward indefinitely.
Most people either haven't turned it on, or turned it on without telling it anything useful.
3 Things Worth Configuring
Every AI memory system, whether it's ChatGPT's Custom Instructions, Claude's memory settings, or a project-based context file, benefits from the same three categories of information. These apply regardless of which tool you use.
1. Who You Are and What You Do
This is the foundation. Your AI needs to know your role, your business model, and the people you serve.
For a fractional CTO, that might look like: "I run a fractional CTO practice serving B2B SaaS companies between $2M and $15M ARR. I advise on technical strategy, team structure, and vendor selection. My clients are usually non-technical founders who need someone to translate between their business goals and their engineering teams."
For a management consultant: "I run a solo consulting practice focused on operational efficiency for professional services firms. My typical engagement is 90 days. I work primarily with firms between 20 and 200 employees who have outgrown their startup processes but aren't ready for enterprise tooling."
Notice the specificity. You're not telling the AI "I'm a consultant." You're giving it the same context you'd give a sharp colleague on their first day. Industry. Client profile. Typical engagement. How you describe what you do.
That context alone eliminates the most common failure mode in AI output. Generic advice written for nobody in particular.
2. How You Work
This is where most people stop short. They tell the AI what they do, but never explain how they prefer to do it.
Preferences and constraints matter. If you write in a direct, conversational tone, say so. If you never use jargon with clients, say that too. If your proposals always follow a specific structure (problem, approach, timeline, investment), include the structure.
Some practical things to include in this layer. Your preferred writing tone, formats you use regularly, technical depth that matches your audience, frameworks or methodologies you rely on, and any specific patterns you want the AI to follow or avoid.
One consultant I work with added a single line to his AI memory: "When I ask for client-facing copy, keep it under 8th grade reading level. When I ask for internal strategy docs, assume the reader has an MBA." That one distinction improved the relevance of every output he got from the tool.
You're essentially building a style guide for your AI. The more precise the guide, the less editing you do on the back end.
3. What You're Working on Right Now
This is the layer most people miss entirely, and the one that has the biggest impact on day-to-day usefulness.
Your AI should know your active clients, your current projects, and the problems you're solving this week. When it has that context, you can say "draft a follow-up email for the Meridian project" instead of spending five minutes explaining who Meridian is, what the project involves, and what happened in the last meeting.
This layer needs regular updates. When you close out a client engagement, remove that context. When you kick off a new project, add it. Think of it the same way you'd think about updating a shared task board. Stale context produces stale output.
In Claude, you can set up separate projects with their own context and memory. In ChatGPT, you can use Projects or Custom GPTs loaded with client-specific information. Either way, the principle is the same. Give the tool enough current context that your prompts can be short and specific.
The difference shows up immediately. A prompt like "suggest three next steps for this engagement" produces generic consulting advice when the AI has no project context. The same prompt, with loaded context about the client, the engagement scope, and the recent deliverables, produces specific recommendations you can actually use.
I tested this with a simple prompt recently. "Write a scope summary for the current engagement." With no context loaded, the AI produced a template with placeholder text. With a project memory that included the client's industry, the engagement goals, and the last three deliverables, it produced a summary I could send with minor edits. Same prompt. Completely different output.
That gap between generic and specific is the gap between AI as a novelty and AI as a genuine productivity tool.
What About Privacy?
This is the question that stops a lot of people from configuring memory at all, and it deserves a straight answer.
Both ChatGPT and Claude give you full control over what gets stored. You can view your memories, edit them, delete individual entries, or wipe everything. Claude offers incognito conversations that bypass memory entirely. ChatGPT has temporary chats that do the same thing.
For client-sensitive work, you have options. You can keep client names out of your memory profile and reference them by project code instead. You can use project-level context (which stays contained) rather than global memory. And if you're on a paid business or enterprise plan, your conversations aren't used for model training by default.
The practical approach is straightforward. Put your working style and general role information in your global memory. Put client-specific details in separate, contained projects. Keep anything genuinely confidential out of both, and reference it by shorthand the AI can recognize from the project context.
You don't have to store everything for memory to be useful. Even a minimal profile, your role, your tone, your typical deliverables, eliminates the cold-start problem that makes AI feel like a waste of time.
The Compounding Effect
AI memory isn't a one-time setup. The context you build compounds. Every correction, every preference you add, every project you load makes the next conversation faster and the output sharper.
Most people evaluate AI based on a cold-start conversation. They type a generic prompt, get a generic response, and conclude the tool isn't ready for real work. That's like hiring a contractor, refusing to brief them, and then complaining about the deliverable.
The difference between "AI doesn't work for me" and "AI saves me five hours a week" usually comes down to that initial setup. Twenty minutes of deliberate configuration changes every conversation that follows.
If you haven't configured your AI memory yet, start with those three categories. Who you are. How you work. What you're working on. Spend 20 minutes on it. The return on that 20 minutes shows up in every conversation you have with the tool from that point forward.
. . .
Want to save hours each week by turning work into repeatable AI workflows?
The Fortune 100 AI Skills Library™ includes plug-and-play prompts built to save leaders time and money. Copy, paste, and edit in 60 seconds, then apply them across planning, execution, and reporting.
