Steve Clean AI and Instructing – The Courageous New Global community

This article beforehand appeared within the Entrepreneur & Innovation Exchange (EIX)

This is the sixteenth 12 months we’ve been instructing the Stanford Lean LaunchPad class. This 12 months, from the primary hour of the primary class, we realized we have been seeing one thing extraordinary occur. It was each the tip and starting of a brand new period. 

Teams confirmed as much as the primary day of sophistication with MVPs (Minimal Viable Products) trying like completed merchandise that earlier lessons had taken weeks or months to construct. After the category, because the instructors sat processing what simply occurred, we realized there’s no going again. 

I’ve been writing about how AI is going to change startups, however the shock of seeing 8 groups really implementing it was thoughts blowing. And not a single staff thought they have been doing something extraordinary.  

---

Class Observations: Product Development Velocity is Off the Scale
The previous sequence for our class was easy – we had groups replicate what they’d do in a startup. Have an concept. Build a staff. Get out of the constructing to speak to clients to know their issues, do Agile improvement and DevSecOps to construct Minimal Viable Products (MVPs) over 10 weeks to check the options. And in the event that they have been going to construct an organization, uncover and  develop a “moat” of proprietary code and options.

This 12 months, within the first week of the category our college students used a number of AI instruments to interchange what beforehand would have taken a big improvement staff. They used Perplexity and ChatGPT for analysis, Claude Code and Replit to construct apps, Vercel/v0 for prototyping, Granola to auto-transcribe and summarize buyer interviews. The complete stream was compressed.

Because it was really easy to have an concept after which construct one thing in minutes/hours, our college students confirmed up on the primary day of the category with merchandise. They now not needed to wait weeks or months earlier than testing whether or not anybody cares.

What we realized we have been watching was a large acceleration of the Customer Discovery / Customer Validation timeline. 

Learning 1. Impedance Mismatch Between Product Development and Learning
By the third week of the category we noticed that the speed of product improvement meant that groups might now generate extra merchandise than they may validate. The quantity of product didn’t equal the quantity of studying. Teams have been so overwhelmed with a lot info from the AI instruments that they overpassed the objective of buyer improvement. They began to imagine that the product itself was the reality.

Consequence 1. AI has made Customer Validation Harder
The abundance and ease of making MVPs has turn into an unintended denial of service attack on the seek for a repeatable and scalable trade mannequin. While that is an artifact of right now, it means we’d like a distinct mannequin for Customer Development as speedy coding isn’t going away.

Learning 2. Student Dependence On ChatGPT Decreased the Quality of Insights After week two of the category, it was clear groups have been delegating communication to an AI. This dumbed down communication became AI slop. ChatGPT and Claude aren’t any substitute for considerate communication – whether or not it’s electronic mail, PowerPoint or weekly summaries of Lessons Learned. Luckily you’ll be able to spot this rapidly.

Learning 3. Customers are Feeling Disrupted
As the scholar groups obtained out of the constructing, they found that potential clients have been already feeling disrupted by AI. Many of the businesses the groups demo’d to realized that they have been seeing not simply incremental enhancements, however in truth have been being proven a “going out of trade” situation.

Learning 4. Customers notice their proprietary knowledge may be their solely moat
In some instances, potential clients who would have beforehand shared their knowledge with college students at the moment are asking for NDAs to share info with the staff. Customers are realizing that carefully held and hard-won info may be one of many few obstacles to AI.

Potential 1: Customer Co-Design
As AI instruments are permitting our groups to construct larger constancy MVPs, a number of are starting to think about using the MVPs as digital twins (as a simulation of the ultimate product.) When put within the cloud and shared with potential earlyvangelists, startups can now begin co-designing the product with potential prospects.

Teams can monitor if the digital twin is getting used, the way it’s used, and the suggestions of what options are wanted could be shared immediately. Teams can replace the digital twin as they add options.

Potential 2: Agent/Customer Outcome Fit
Today, software program functions are constructed to offer customers info after which anticipate the customers to do the work through a consumer interface of dashboards, alerts, workflow instruments and stories. But clients purchase software program to get a job finished, not to have a look at extra screens. Getting the job finished is what AI Agents (orchestrated by instruments like OpenClaw) will autonomously allow. For some groups, future class sections may even see the seek for Product/Market match turn into the seek for AI Agent/Customer Outcome match. Minimum Viable Products (MVPs) will turn into Minimum Productive Outcomes (MPOs.)

Lessons Learned

• MVPs are No Longer an Indication of Technical Competence
  • Vibe coding has remodeled MVPs to the equal of PowerPoint slides
• Speed to MVPs Hasn’t Yet Meant Faster Learning About Building a Company
  • While we’re nonetheless early within the class, the blinding pace of the primary week’s onslaught of MVPs hasn’t but translated into sooner studying about buyer validation.
• Business Process and Business Models Still Matter
  • The bottleneck for our scholar groups has moved from needing the sources to construct high-quality MVPs to judgment: how to decide on the suitable drawback, find out how to learn consumer alerts appropriately, and deciding what to construct subsequent.
• Product/Market Fit and Agent/Outcome Fit Will Co-Exist (for some time.)
  • While some clients are prepared to maneuver to an Agentic workflow, for others delivering Product/Market Fit continues to be what customers need to see.
• Startup Teams Will Be Smaller
  • Our class groups are 4-5. In the previous, in the event that they determined to pursue their concept and begin an organization they would wish to rent a bigger staff to construct the product, handle the product, discover out whether or not they had product/market match, create demand, and many others. That’s largely now not true.
  • Most groups gained’t want to boost cash to search out out if the issue is actual or earlier than they know if customers care.
• Enterprise Pricing Models Will Change
  • Some groups are already testing pricing that may shift from per/seat to workflows, outcomes, outcomes, resolutions, profitable activity
• Customer Development Will Change
  • Because the Customer Development cycle is quicker and a number of MVPs now could be run concurrently…
  • Effort shifts to the additional time wanted on hypotheses testing as a result of the speed and quantity of product improvement can overwhelm alerts from potential clients
  • As MVPs quickly change, they must be instrumented to observe buyer utilization/interactions

More Learning In the Weeks Ahead

Filed underneath: Lean LaunchPad, Technology |