Richa Deo
I spent years researching why technology adoption fails. Now I build AI products that test those findings in practice.
Built enterprise AI agents, AI diagnostics, and adoption experiments used by employees, teams, and businesses across multiple countries.
Organic reach
India, France, and the United States — no paid marketing (FoundIndex)
275+ employees
Trained in practical AI agent creation
2,000+ employees
Applicable on production approval (Feature Builder)
9/10 user rating
Across 35+ testing sessions
Projects
Three current betsEnterprise AI Agent · BT Group
Feature Builder
- Problem
- Enterprise teams needed to create compliant work items but the process was too complex. Most people skipped compliance rather than engage with it.
- Solution
- Built Feature Builder using Microsoft Copilot Agent Builder. The agent guides users through requirements conversationally, removing friction at point of authoring.
- Results
- 50+ pilot users
- 9/10 user rating across 35+ testing sessions
- Navigating formal AI governance
- Applicable to 2,000+ employees on approval
AI Visibility Diagnostic Tool · Live Product
FoundIndex
- Problem
- Businesses had no way to see how AI systems interpret their websites.
- Solution
- Built FoundIndex using Lovable, Supabase, and OpenAI. Users enter their website URL and receive a diagnostic showing how AI systems read their structure and content.
- Results
- 397 active users at peak
- 155 active users last 30 days
- Users from India, France, and United States
- No paid marketing
- 1 minute average engagement at peak
AI Prototype · Candidate Experience
FoundCandidate
- Problem
- Hiring processes create uncertainty for candidates.
- Hypothesis
- AI can reduce ambiguity without adding recruiter workload.
- Current status
- Prototype. Testing assumptions about where candidates experience the most uncertainty.
Product Experiments
Hypothesis → Test → Learning → Next Test
| Experiment | Hypothesis | Outcome | Next Test |
|---|
| FoundIndex | Users will use an AI visibility diagnostic to improve their websites | Users wanted to understand AI visibility more than act on optimisations immediately | What monetization model best aligns with user value? |
| Feature Builder | Removing friction at point of authoring will increase compliance adoption | Friction — not unwillingness — was the real barrier | Will removing authoring friction increase compliance scores at scale? |
| FoundCandidate | AI can reduce candidate uncertainty during hiring | Prototype revealed trust concerns not visible in spec documents | Does transparency at each hiring stage reduce candidate dropout? |
What I Have Learned About AI Adoption
Finding 1
Trust beats usability
People do not avoid AI tools because they are hard to use. They avoid them because they do not trust the output enough to act on it.
Finding 2
Capability gaps masquerade as product problems
When adoption fails, organisations blame the product. Usually the real barrier is that people do not know what to do with the output once they have it.
Finding 3
Adoption happens when uncertainty becomes manageable
The moment a person feels confident enough to take the next step — not expert, just confident — adoption follows. Products that create that moment win.
How I Build
01
Identify behaviour problem
02
Form hypothesis
03
Build prototype
04
Test with users
05
Measure adoption
06
Iterate
Tools I Build With
Microsoft Copilot Agent BuilderLovableSupabaseOpenAI APIClaudeKiroPrompt EngineeringAI Agent DesignGoogle Analytics
Current Questions I'm Exploring
- →How do AI agents earn trust in enterprise environments?
- →What causes users to return after first AI tool usage?
- →How much friction must be removed before compliance becomes adoption?
- →What signals tell users an AI output is safe to act upon?