There comes a point in every marketing, sales, and business development organization where the demand for internal knowledge begins to outpace the capacity to deliver it. The volume of content grows, the number of use cases expands, and the expectation for fast, accurate answers increases daily.
And systems like SharePoint, despite being exceptional at housing information, have a remarkable ability to hide it at the exact moment you need it most.
This wasn’t a new challenge for us.
As our services evolved, so did the proliferation of decks, case studies, proposals, industry insights, and versions thereof - dispersed across multiple SharePoint sites, nested folders, archives, and legacy structures. Add the complexity of each seller needing something different depending on their vertical, prospect maturity, or the specific conversation they’re preparing for, and the problem magnified.
As an AI ambassador in our organization, I had already completed advanced programs through MIT Sloan, MIT CSAIL, AWS, and various industry-led GenAI courses. I had experience guiding teams through AI adoption and emerging familiarity with concepts like retrieval-augmented generation (RAG), vectorization, grounding, and the strategic use of unstructured data.
But this challenge wasn’t theoretical.
It was operational, immediate, and deeply tied to the efforts of our revenue teams.
It demanded a solution that didn’t just automate tasks.
It needed to meaningfully help people do their jobs better and faster.
Before opening Microsoft Copilot Studio, I returned to the mindset shaped by years of digital strategy, internal IT leadership, sales enablement, and content operations:
The right question wasn’t, “What can AI do?”
It was, “What would genuinely help our teams move faster?”
Sales and BD teams don’t need more content.
They need the right content surfaced instantly, contextualized to their audience, and delivered in a format they can act on immediately.
They need:
For me, this raised an additional requirement:
How could I introduce this system without adding another tool to their already crowded ecosystem?
That’s where Copilot Studio, combined with Microsoft Teams integration, created the breakthrough.
My exposure to GenAI, RAG, and applied AI strategy laid the foundation, but building something truly useful required shifting from theoretical knowledge to practical design.
I approached Copilot Studio not as a developer, but from a no-code/low-code AI strategy perspective.
What immediately impressed me was how its no-code environment allowed me to create a workflow-driven experience shaped entirely around how our sales and BD teams actually operate. With it, I could:
And critically:
Licensing constraints meant we needed a scalable way to deliver this solution without onboarding dozens of new seats.
The answer was the Teams connector. With it, the system didn’t become another destination.
It became part of where our teams already lived.
The finished application became more than a bot.
It evolved into a knowledge companion I named Onit - a nod to our company name, and a phrase I often used when team members would ask me to tackle deep research: “I’m on it!”
Here is what Onit delivers:
It pulls from numerous SharePoint collections: service frameworks, case studies, industry collateral, decks, competitive intel, and discovery materials. No more guessing which site holds which asset.
Every answer is anchored in real documents.
No internet drift.
Reduced pattern-generated fluff.
Just enterprise knowledge retrieval applied responsibly.
A seller can ask things like:
Onit interprets the intent, classifies the request, retrieves the right assets, and synthesizes the answer with contextual relevance.
Each output includes:
This keeps human-in-the-loop accountability intact and allows for situational accuracy where older documents may still exist.
Which means:
Before rolling Onit out more broadly, I invited a group of cross-functional stakeholders to pressure-test the system. Although the primary audience was Sales and Business Development, I requested feedback from early testers across the organization - including Consultants, our Director of Data Services, our Managing Director of AI, and our CTO.
Their involvement, feedback, and enthusiasm proved invaluable.
They approached the tool with different assumptions, different terminology, and different expectations of accuracy, which exposed edge cases and strengthened the retrieval logic.
Their feedback ensured Onit wasn’t just a sales enablement assistant, but an enterprise-ready knowledge system capable of supporting broader teams as our adoption evolves.
The importance of verifying any info from any model remains a best practice – due to their probabilistic pattern-seeking behaviour, there are times where Onit can still provide info that is close to the desired outcome, but not accurate enough to send to a client or prospect.
As our CTO, Tim Siemens, recently articulated:
“AI is an extraordinary tool, but tools don’t have judgment, integrity, or the ability to look a client in the eye and give unflinching advice. That combination – world-class humans plus world-class AI – is an extremely wide moat. Most companies will get the AI part eventually, but very few will keep the human part this strong.”
This thinking remains central to how Onit operates.
AI accelerates - but humans validate, refine, and ensure precision.
Sales and BD teams who previously spent 10–20 minutes digging through SharePoint now had answers in seconds. The time saved moved directly into higher-value activities: refining messaging, tailoring proposals, preparing more strategically, and elevating conversations.
More importantly, expectations shifted. Teams began asking more advanced, strategic questions - because they trusted that the system could support them.
This wasn’t just about speed. It was about raising the floor and the ceiling of the work.
Looking back, the progression feels natural: studying GenAI academically, applying it operationally, advising our teams, and ultimately building a real no-code, Teams-embedded RAG application that changed the rhythm of our business.
GenAI remains one of the most powerful democratizers of capability.
And platforms like Copilot Studio, with their no-code accessibility, make it possible for domain experts - not just developers - to build meaningful AI applications.
What began as a recurring frustration turned into a system that now supports multiple departments, accelerates quality of work, and opens the door to future client-facing and internal AI innovations.
A simple problem.
A strategic application of AI.
And a solution that has become part of our daily operating fabric.
About the Author
Jamie L. Michie is a dynamic communications leader with 20 years of driving digital initiatives for clients. She blends analytical thinking with creative problem-solving to implement effective solutions to key stakeholders, such as OBS Global's Innovation Lab, and the Data, GenAI, and Financial Services teams.
A lifelong learner, Jamie has recently earned certifications from MIT CSAIL in AI Strategy, and MIT Sloan for Algorithmic Business Thinking, Business Process Design, and GenAI Essentials from AWS.