By Mark Clancy & Linh Phan, Co-Founders of Khashmere, an elite AI and data engineering service that can be resold by MSPs to their enterprise customers.
You see anxiety-driven headlines like “MSPs face growing pressure as AI accelerates threats” (source). You feel intense pressure to create a new revenue stream for your MSP as your customers ask how to use AI in their organizations. If you are not the go-to strategic advisor on AI, you will lose your role as THE tech advisor. They will find AI expertise elsewhere.
A Chatbot With Your Logo On It Isn’t AI Strategy
Your immediate response to the existential threat of “AI [becoming] an MSP competitive risk” (source) is adding Copilot licenses and using prompt engineering to adjust your AI chatbot’s responses – a chatbot that has no access to your customers’ data, systems and processes. You call that “AI.”
Chatbots are only 5% of the revenue opportunity. 95% of what AI can actually do lies in something else: agentic orchestration.
Unlike workflows, which are rigid and structured – think of Power Automate or Zapier – agentic orchestration lets AI agents reason, plan, and self-correct. Your customers’ data is scattered across Power BI reports and QuickBooks invoices. Orchestrated agents reason and act across all of those systems.
We Turned One Client’s Chaotic Inbox Into a Self-Sorting Pipeline
Our client had no way of knowing when they received a sales lead. To further complicate the matter, customer support requests were mixed in with sales leads bloating their inbox.
Our MSP client wanted a more qualified sales lead pipeline while still keeping the “Contact Us” form on their website easy to use. Rather than just capturing the lead and generating an alert, our client wanted AI to triage, qualify, and enrich each lead.
Once we understood the client’s goals, we structured the process and got started. The agentic AI process we designed had three distinct phases:
1. Capture
First, we captured the submitted information in a secure, monitored data store. In this case, we ingested data from a WordPress site via Gravity Forms then used Microsoft Fabric and something called an Eventstream to capture the lead data in real-time.
2. Triage & Enrichment
Next, we called an AI model – which also was not the newest model – Claude Haiku 4.5, to triage and enrich the data.
The triage step identified whether this truly was a sales lead or a customer mistakenly using the form to contact customer service.
The enrichment step involved several tasks:
* Scoring the lead (i.e. probability of success)
* Researching the size of the opportunity (i.e. revenue, number of employees)
* Identifying key personnel (i.e. LinkedIn details of key executives)
* Providing a detailed overview of the sales lead
3. Notifications & Alerts
Finally, we developed notifications and alerting systems in Microsoft Teams channel and updated the sales lead tracking sheet in Microsoft Excel.
Notifications kept key members of the sales and executive teams aware of the lead and its details.
Alerts escalated any errors encountered during the process to IT.
So how does this differ from automating a workflow?
The real magic here is something simply called “context.”
Think of context as multiple sets of business rules that the agent has access to orchestrate the workflow by the AI model. This context will become the “secret sauce” of any business since it forms the foundation of agentic processes that will run the business.
This process is called context engineering, a term Anthropic has written about – read more here if you want to dive deeper.
Now our client has sales leads enriched and scored, and customer support tickets routed immediately to the right place.
Both get flagged, score, and triaged the second they land.
This saves our client hours of sifting and categorizing each week, freeing the team to close deals.
Anyone Can Resell Copilot. Only You Know Your Customer
Remember the role you were afraid of losing?
This is how you keep it.
Any competitor can resell Copilot licenses and embed AI chatbots with defined prompts.
You’ve been a “Managed” Service Provider for your customers for years – managing their tickets, supporting their infrastructure, and putting out their fires.
What competitors can’t replicate is the knowledge of your customers’ data, systems and processes that form the foundation of their business.
That’s your moat.
Now you run agentic orchestration on a team of agents that does the grunt work – sorting, scoring, routing, updating.
You monitor these key processes and intervene to make the final judgment calls when the AI agent can’t.
This is what the “M” stands for in MSP.
This is the type of service that adds high-margin revenue to your MSP’s business.
Find the Task That Eats Their Week. Automate It.
The first step is to work with your customer on an AI readiness assessment.
Find the workflow that takes a lot of their time, then orchestrate it.
That is where you learn their data, systems, and rules to build the context you need.
That’s how you stay the go-to advisor instead of watching someone else take the seat.
“Dream big about your business. We’re the engineers who make it real — collaborating with you, building agentic AI that moves your numbers.”
