Stop Planning Your Agent Strategy and Start Shipping Agents
Build evidence and momentum, then focus on a transformation plan.
TLDR; Enterprises winning with AI agents aren’t the ones with perfect strategies—they’re the ones shipping focused agents, learning fast, and scaling. Start small, solve one painful problem end-to-end, and iterate.
Brett Taylor, who's built society changing products (multiple times – Google Maps, the Like Button, etc etc), shared a perspective recently that perfectly captures what we're seeing with enterprise clients: (worth listening to his recent podcast episode of Lenny’s Podcast)
"I think agents will truly start to bend the curve again like we did in the very early days of computing."
He's comparing this to the 90s productivity explosion when ERP systems first digitized business operations. Not incremental improvements—fundamental transformation.
Yet many enterprises are stuck in "comprehensive strategy" mode while nimble competitors deploy agent after agent and learn from real-world performance.
The Planning Trap
Most enterprises spend months developing "comprehensive agent strategies" while competitors deploy focused agents and learn from real performance.
The pattern we see working
Instead of boiling the ocean, successful companies launch focused agents that handle specific jobs completely:
Resolution agents that handle high-volume customer issues end-to-end
Scheduling agents that manage complex resource coordination
Verification agents that process routine approvals in real-time
Fulfillment agents that manage standard requests from intake to completion
Each agent owns one specific job end-to-end. This mirrors the parallel that Taylor described about mechanical engineering firms during their transformative year: entire drafting departments didn't get reduced—they ultimately got eliminated. But that change happened one specific part of their job at a time.
What Actually Works
Looking at enterprises winning with agents versus those stuck in pilots, clear patterns emerge:
Winners focus on complete jobs, not AI assistants. Instead of "help with customer inquiries," they build agents that "resolve billing questions from first contact to resolution."
Winners embed domain expertise. Your best customer service rep should help design the service agent. Your top operations manager should architect the scheduling agent. Technical teams enable; domain experts lead.
Winners ship and learn. Launch in weeks, not months. Gather performance data daily. Iterate based on real interactions. If you never ship, you never learn.. It sounds ridiculous to type this, but we see too many orgs stuck in discussion about what to do. Winners are reviewing what’s happening and quickly doing what they can to make it better.
Winners think beyond efficiency. What new services could you offer with instant response times? What markets could you enter if expertise scaled infinitely? It’s ok to start with efficiency AI, but don’t leave growth AI off the table. Some of the most interesting work we’re seeing is happening with orgs that aren’t thinking beyond efficiency.
Let me be clear though, efficient AI is the easy (and safer) place to start. Every modern organization trying to do anything with AI should be constantly experimenting with how agents can help automate, simplify, or even eliminate inefficient parts of work.
If you’re not doing this yet, please call us :)
A 60-day Framework to Get You Started
Week 1: Pick Your Most Painful Job Identify the one interaction that happens constantly, follows predictable patterns, requires expertise but not crazy creativity, and currently burns expensive human time. Get clear about the problem to solve, document what you can about how the solution works today, try to understand the time/cost with the current state.
Week 2-3: Build WITH Domain Experts Put your best practitioner(s) in the room with actual builders. Not consulting. Not reviewing. But building together. The measure of their success is simple; did something get built?
Week 4-6: Ship Something Imperfect Launch criteria: handles 70% of cases correctly, escalates appropriately when uncertain, logs everything for improvement.
Week 6-8: Scale What Works Double down on what's exceeding expectations. Kill what isn't delivering value. Use key learnings to design the next agent. Step, repeat.
The Technical Reality
Everything about how we build agents will be different in six months. The models will improve. The tools will evolve. The techniques will advance.
But the customer’s need for problem-solving is permanent.
Hold your problems sacred. Hold everything else—the tech stack, the model choice, the prompting approach—lightly.
The Competitive Clock
Taylor's warning should galvanize every enterprise leader: "The window for first-mover advantage is measured in weeks or months, not years."
Every week in committee is a week your competitors spend learning from agents in production. They're gathering performance data, refining approaches, and building institutional knowledge that compounds daily.
Meanwhile, enterprises stuck in planning mode are falling further behind in the most important race: learning how to make agents work in their specific context.
Questions for Your Next Leadership Meeting
Instead of asking "What's our AI strategy?" ask your team:
What's the most painful, repetitive job we handle daily?
Can we move fast enough with focused agents while competitors plan comprehensive strategies?
Will we empower our best practitioners to lead agent development?
Are we ready to scale what works aggressively?
The Choice
The enterprises (we see) that are thriving with agents aren't the ones with the best plans. They're the ones who understood that agents had the potential to change everything. And they were willing to take a risk for the sake of learning.
Your domain expertise is your competitive advantage. But only if you use it to build experiences (agents) that work, not strategies that sit in slide decks.
The question isn't whether agents will transform your operations. They will.
The question is whether you'll lead that transformation by shipping real technology, or watch competitors do it while you're still debating use cases.
Ready to move from planning to shipping? The most successful enterprises are launching focused agents in 90 days, not comprehensive strategies in 12 months.
 | A guest post by
|