Most AI pilots die in the pilot stage — not because the technology fails, but because nobody defined what it was supposed to fix. On the latest episode of the CanadianSME Small Business Podcast, host Maheen Bari sat down with Kishore Kannan, CEO and Co-Founder of RAGWorks AI, to unpack how agentic AI is rewriting sales pipelines and back-office workflows — and why small businesses may be better positioned to exploit it than the enterprises they compete with.
The End of the Manual SDR
Business development has traditionally run on human hours: building lead lists, personalizing messages, choosing channels, sending emails. Kannan argues almost none of that needs a person anymore. “The major chunk of their day-to-day activities — which is research and reaching out — could be all automated and executed with AI agents,” he said. His pitch to clients is blunt: “You don’t need a five-person sales team just to do the outreach.”
But volume is not the point; timing is. Prospects raise seed rounds, open hiring sprees, change direction. “The idea is to capture the right intent and reach that lead at the right moment,” Kannan said. “That adds the value.” Prospecting becomes a background utility that runs around the clock, leaving human teams to do the one thing agents cannot: close.
Why Pilots Die — and How to Keep Yours Alive
Ask Kannan why so many companies stall in pilot mode and he flips the premise. Starting with “I want to adopt AI” is starting backwards, he said. Begin with a core problem, then ask what automation can do about it.
Much of the paralysis, he suggested, is fear of being left behind — companies jumping in without a defined problem, then quietly giving up when nothing measurable happens. His prescription is narrow: “Start with a very specific problem in mind, start small, and then keep expanding the agent scope.”
Do not automate sales; automate one slice of outreach, measure the outcome, then hand off to the next agent — or to a human, where trust and relationships close the deal. “Measuring the outcome and breaking down the scope is really crucial,” he said. Companies that skip that step “eventually end up giving up.”
Building an Agentic Workforce, Department by Department
Every department is sitting on a workflow that eats hours: finance has month-end reconciliation; support has an inbox of tickets that each require digging through knowledge bases. The approach is the same — scope the workflow, define the outcome, connect the agent to the right tools.
That means plugging agents into systems the business already runs — the ERP for invoicing, the CRM for pipeline — building them one at a time, each with a dedicated purpose. Then comes orchestration: an agent that reads the inbox, decides whether a message is an invoice or a support issue, and routes it to the right specialist. RAGWorks scaled that way with clients — start with outreach, and once results are visible, requests for back-office automation follow.
The Cost Nobody Budgets For
The most common misconception, in Kannan’s view, is the belief that AI solves everything — and the shiny-tool reflex that follows. A company spins up a chatbot on its website data, a hundred users ask a hundred different questions, hallucinations appear, and the verdict is that the pilot failed.
The second blind spot is money. Agents run on tokens, and tokens are real cost — he pointed to coding agents left running unattended, generating entire ecosystems nobody asked for. Before deploying, an organization should know two numbers: what the agent is expected to deliver, and what it will cost to run. “Scoping it down, considering cost and performance, would be the key.”
That includes model choice. “People say Claude can do everything, or OpenAI can do everything,” he said, “but you can literally find differences between these models.” Image extraction may be a job for Gemini. The expertise is in matching the model to the outcome, not defaulting to the loudest brand.
Why Small Businesses Have the Advantage
Here Kannan is emphatic: the small operator is better placed than the enterprise. Large organizations are weighed down by legacy tools, change-management processes and approval policies. “By the time it gets in, a lot of advancements have already happened,” he said.
Small businesses can test a tool, see the outcome quickly, and adopt it. RAGWorks is its own proof: it built an AI SDR for internal use before selling it to anyone. “We were the first customers for ourselves,” Kannan said. There was no sales team — agents did the outreach, booked the meetings, produced the conversions.
The Bottom Line: Map the Bottleneck, Not the News Cycle
For entrepreneurs overwhelmed by the pace of releases, Kannan’s advice is to stop reading and start diagnosing. Identify what is consuming the most time in the business, and map that single problem to a single agent. Experiment with one or two models and measure the result.
His own company began exactly there: the core problem was finding the right sales hire, and the interim solution was an agent that got people talking about the business. Everything else was built on top of that. The lesson for Canadian SMBs is not that AI is optional — it is that the broken process, not the trendy model, is where the work starts.
To learn more, visit canadiansme.ca, and visit the CanadianSME Small Business Foundation at smbfoundation.ca. A special thank-you to podcast partners UPS and ADP for their ongoing commitment to empowering small businesses across Canada.
Disclaimer: This article is based on publicly available information intended only for informational purposes. CanadianSME Small Business Magazine does not endorse or guarantee any products or services mentioned. Readers are advised to conduct their research and due diligence before making business decisions.
