⬛️ playbook: semi-autonomous innovation

playbook last updated feb 6th 2025

building a new business? today you can get from discovery to validation faster than ever. this of course isn’t a perfect approach (innovation never is, it’s a subtle combination of vision, timing, execution and luck!) but it’s the best i’ve found so far.

i'll continuously be tweaking this and sharing insights along the way

(disclaimer: yes we still need a human in the loop, ‘good taste’ is more important than ever, co-intelligence is the way to go, but for how long?)

ok so…here’s how i’d do it (and am doing it) in 2025, using ai to handle 80% of the work:

a now tried and tested playbook

(note: no time to read, here’s an audio version you can list to)

The playbook

here’s a high-rez pdf version of the image you can download 👇

step 1: identify opportunities

this step usually doesn’t happen from scratch, but i have some solutions if you’re totally out of inspiration

→ tap into your “secrets” (peter thiel)

opportunities typically stem from customer pain points, or spotting inefficiencies 

if you’re in a specific industry or space, you’ll probably have spotted these pain points or inefficiencies, your job will then be to test whether this is a pain worth solving + whether someone will pay to solve it (more on that later)

→ if you’re starting from a blank page you can use these tools to uncover up-and-coming opportunities and/or validate yours

 treendly

glimpse (which has great segmentation criteria)

both are valuable to spot untapped niches and promising ideas.

and finally google trends (which you already know)

bonus: and of course, to validate these ideas you have semrush, ahrefs, ubersuggest 

step 2: empathize with users / ideate solutions

this is where things get really interesting.

approaches like the mom test, the 5-whys approach or the jobs to be done framework allow you to go to the core of a pain, the needs, the benefits, the jobs to be done. these are tried and tested.

however running these and empathising with users typically takes weeks, months of customer interviews. it’s so time consuming.

so…introducing ‘synthetic data and synthetic user interviews’

synthetic data and customer interviews are created using ai, they replicate human interactions. they are artificially generated data, digital clones of people, that mimic real-world data but does not contain any actual personal information

the good news is that this drastically reduces costs and time to insights

and yes, organic interviews capture real-time, nuanced human feedback, while synthetic interviews use ai-generated responses to simulate diverse user interactions, so you’re getting speed and scalability but potentially missing subtle emotional and contextual cues

speed vs perfection? you decide

while these won’t replace real-world interviews (or can they?), they’ll definitely complement them

please please please test it out before judging, you’ll be amazed

for this i’d suggest 2 tools

→this ‘user persona generator’ (it’s a public customgpt)

i have conversations with these regularly, here i’m asking about price sensitivity

here’s an example of the output

and in the synthetic user’s own words:

"I’ve been leading marketing campaigns for years, and AI seems like a game-changer—but honestly, it’s overwhelming. There are too many tools, and I don’t have time to figure out which ones actually work together. My boss wants more personalized outreach and better lead conversion, but without adding more workload to my already stretched team.

I need a straightforward AI playbook that shows me exactly what to automate, which tools to use, and how to integrate them without needing an IT degree. If I could set up AI workflows that improve efficiency without making marketing feel robotic, that would be a game-changer."

here’s a second persona:

try it out yourself here:

next up…. if you’re willing to invest a bit of money for better output

→ try out syntheticusers to uncover real pain points

well worth the price. you can have back and forth synthetic interactions with your customers. this is god-send.

here’s the pricing

try this out

→ interact until you gather actionable solution concepts.

then if you want real customer interactions: linkedin or respondent

ok next step…

step 3: prototype solutions

now it’s time to build your prototype, your mvp, or you mdp (minimal delightful product) or whatever you wanna call it

this used to take tooooons of time and effort…briefings, find a designer, more briefings, paper prototype etc etc…lose patience, lose momentum

but prototyping just became much much easier

(note: no time to go through all these solutions i’m suggesting? simply go to bolt.new and play around with it)

if you have a bit more time, here’s what i recommend:

→ for low-fi prototypes

simple slide deck: use gamma and/or slidespilot

for an ai-generated landing page: relume or framer

→ for mid-fi prototypes:

create quick prototypes with bolt, lovable or replit

here i’m a prototyping 3 ideas at the same time (feel like an prototyping god)

all clickable, all deployed… spend a few hours on this.

→ for hi-fi prototypes:

you’ll need a lot of patience and/or help from a dev here

recommended tools:

 cursor, claude artifacts, or github copilot to build fast, usable demos.

next step?

step 4: pre-optimize with synthetic feedback

→ gather early feedback using wevo and/or syntheticusers before major resource commitments.

wevo is cool as it offers a score, some optimisation ideas for your landing page and/or prototype (it will need to be deployed though)

→ refine and iterate swiftly based on findings

step 5: create outreach strategy

so here you have 2 approaches

1. brute force it using llms. this all comes down to some good prompting on gpt4o, r1, claude or perplexity)…”i need the perfect early adopter outreach strategy for x, y, z” then refine

or

2. a little gem that allows you to define channels and messaging: m1-project 

you can thank me later for this one, it’s pure gold

note: they’re friends, use this promo code for 25% off: XXX25

this will generate your business overview, icps, channel ideas, pricing experiments, marketing tactics and more

not perfect, but…if you’re not a marketing pro, this is great!

have fun!

step 6: start validation experiments

validation experiments allow you to validate your hypotheses

learn more about the different types of experiments here

the channels you’ll use to attract your first prospects/ideal customers will really really depend on the output of your outreach strategy in step 5

i’ll write more about this soon, with a full playbook

in the meantime here are some basic ideas

→ generate and recylcle targeted content with newtn or writesonic
→ reach out directly via lemlist
→ got some budget… test ads with adcreative.ai plus koast if you’re ready to scale.

for a full validation roadmap i also love this approach:

go to perplexity
switch on ‘pro’ and reasoning R1

use this prompt:
“I'm launching a [insert your business idea in as much detail as possible] please help me analyze market fundamentals, research competition, revenue models and create my validation framework”

here’s some of the output:

keep tinkering…

→ gather results, iterate, and amplify.

what’s next?

rinse and repeat: go back to step 3 until you feel you reach validation

remember, your riskiest assumption is usually ‘willingness to pay’

here are a bunch of others

the result of all this?

one person becomes a fully operational team:
this technology is giving us wings.

disclaimers!

→ this is not a perfect formula, but it’s working for me

→ you test, learn, and refine at blazing speed (days not months)

→ you're still the human in the loop, this is co-intelligence

→ synthetic data first then validate with real humans

→ ´taste’, eval and intuition matter more than ever

→ test faster, iterate rapidly

→ you spend more time on insights than on execution

→ for the build phase, you still need a dev (a 10x dev)

you don’t need a huge team or budget to make an impact, just the right tools and mindset. fun to see because it’s already happening.

more coming soon!

⬛️ david

ps: some previous and relevant playbooks/posts you might have missed