Many teams have a Notion page full of prompts. Very few have real, repeatable AI workflows. In this episode, host Susan Diaz and product/go-to-market leader Jason Dea dig into how to move from playing with prompts to designing workflows, building tiny specialist agents, and avoiding a new wave of shadow AI inside organizations.

Susan is joined by venture studio and SaaS veteran Jason Dea from Coru Ventures in Toronto.

They unpack why AI is not a magic wand or a single feature, but an enabling technology that only delivers value when it's wired into actual workflows.

Jason shares his "swarm of bumblebees" metaphor for AI, how he builds small specialist agents to clone his own work style, and why enterprises are about to repeat the mistakes of shadow IT if they don't get serious about orchestration and governance.

They close by talking about leaders using AI in their own day-to-day work, and Jason's personal experiments with family apps, coding, and even a butterfly-catching game for his daughter.

Key takeaways

Prompts ≠ workflows. Collecting prompts in a shared doc feels productive. But until you map the 8–10 steps of a job and decide where AI fits, you're just doing experiments, not transformation.

AI is not a magic one-shot. It's an enabling technology. The real gains come when you see your work as a chain of small tasks and let AI take over the repetitive, boring, or "toil" links in that chain.

Think "swarm of bumblebees." You are the queen bee. AI is a swarm of tiny worker bees, each doing one specific task very well (emails, slides, requirements, research), not one mega-agent doing everything.

Documenting workflows doesn't have to be fancy. A workflow is just "tell me the 10 steps." Start with the human sequence. Tools come second. Once it's visible, the friction points where AI can help become obvious.

Shadow IT is turning into shadow AI. Cheap, bolt-on AI features and swipe-a-card tools make it easy for every team to spin up their own stack. Without orchestration, you recreate silos, risk, and tool sprawl at AI speed.

IT should govern, not own everything. Governance, security, and guardrails matter. But AI also democratises small bits of "coding" and automation, letting non-technical teams build more, faster—if they have guidance.

Leaders need hands-on literacy. The fastest way out of the hype is to use AI yourself for your own toil. Drafting emails. Planning. Decomposing big tasks. You get more realistic about what it can and cannot do.

AI is an "unstuck" tool in work and life. From relearning to code, to building tiny family apps, to cataloguing knick-knacks and designing games for kids, AI opens up projects that were unrealistic even five years ago.

Episode highlights

[00:01 ] Jason's background in startups, SaaS, product, and go-to-market, and his role at Coru Ventures.

[02:00 ] Where we are on the Gartner hype cycle and why the trough of disillusionment is inevitable and useful.

[04:40 ] Why some people can't imagine life before ChatGPT—and why that's not true for everyone inside organisations.

[05:50 ] Mapping work as a sequence of steps instead of hunting for a single "magic" AI prompt.

[08:01 ] The "swarm of bumblebees" metaphor: you as the queen, AI as many small worker-bee agents.

[09:59 ] How to define a workflow in plain language: "tell me the 10 steps," tools aside.

[11:00 ] Paperwork and OCR as a classic example of where generative AI finally unlocks messy, grey-area tasks.

[13:50 ] Using AI first to remove the tasks you hate and identify the links you should outsource to machines.

[15:20 ] Jason's "digital clone" AIs trained on his own content and patterns.

[19:00 ] Building multiple mini-AIs: one for social posts, one for slide decks, one for product requirements.

[21:10 ] Bolt-on AI features everywhere + messy workflows = amplified confusion and risk.

[22:10 ] From shadow IT to shadow AI: why orchestration and shared understanding of workflows is critical.

[24:40 ] Startups' speed vs enterprises' risk aversion, and what each can learn from the other.

[27:10 ] Why IT should set guardrails while letting departments experiment and build more on their own.

[30:10 ] Jason's advice to leaders: use AI yourself to see where it really helps and what it really takes.

[36:00 ] Personal-life AI: relearning to code, family apps, cataloguing home items, and a butterfly game for his daughter.

[38:00 ] Susan's idea: vibe-coding a family recipe app as a way to preserve memories and workflows.

If your organization has a folder full of prompts but no clear AI workflows, this episode is your sign to pause and rethink.

Share it with:

• The person who keeps buying new AI tools.

• The leader who thinks "IT will figure it out".

• The teammate who's already acting like the queen bee and quietly building their own swarm.

Then ask as a team: "Where are our 10-step workflows, and which links should really be done by AI?"

Connect with Susan Diaz on LinkedIn to get a conversation started.

Agile teams move fast. Grab our 10 AI Deep Research Prompts to see how proven frameworks can unlock clarity in hours, not months. Find the prompt pack here.