AI UX consultancy · Powered by Groundwork
Build the AI that's worth building.
And skip the rest. We check your AI idea, launch, or vendor against real failure patterns. Then we tell you what to fix first, or not to build it at all.
Maps findings to the EU AI Act and NIST AI RMF. Flags issues for legal review. Never certifies.
What you bring
- A use case
AI triage for support tickets
- A launch plan
Pilot with two teams
- An AI vendor
Not chosen yet
Groundwork
- 01Use case readiness
- 02Launch risk
- 03Platform risk
- 04Automation tracesoon
- 05Agent groundingsoon
- 06UX auditsoon
What you get
Verdict
RiskPatterns behind it
- Validated user need
- Workflow impact
- Uncertainty handling
Fix first
Show agents why a ticket was routed. Let them correct it in one click.
Groundwork does the diagnosis. Nelly reviews it. You make the call.
Why now
Most AI features don't fail on the model. They fail around it.
Teams rush to ship AI. The real problems show up later: in the workflow, in trust, in compliance. By then they're expensive to fix.
- Is it solving a real user need?
- What happens when it's wrong?
- Who is accountable?
- Will it pass the EU AI Act?
[TODO: stat]
AI pilots that never reach production
[TODO: stat + source]
[TODO: stat]
AI projects abandoned after proof of concept
[TODO: stat + source]
[TODO: stat]
Users who stop trusting AI after one bad answer
[TODO: stat + source]
[TODO: stat]
Cost of fixing a problem after launch vs. before
[TODO: stat + source]
The framework
Six ways AI products break.
From my book, The Interface Is Not the System. Every Groundwork report scores against these.
- 01
Feature vs. strategy
AI gets bolted on as a feature, with no link to what the business needs to win.
- 02
Accuracy vs. usefulness
A model can be accurate and still not help anyone do their job.
- 03
Uncertainty handling
The product hides when the AI is unsure. So users trust it at the wrong moments.
- 04
Workflow impact
The AI works. The work around it gets slower, messier, or riskier.
- 05
UX timing
The AI shows up at the wrong step. Too early or too late to change a decision.
- 06
Validated user need
Nobody checked that users want it. The demo did the convincing.
How it works
Groundwork does the heavy lifting. You make the final call.
01 · Scan
Tell Groundwork what you're evaluating. It routes it to the right diagnostic.
02 · Review
I review the report with you. Evidence first, no hype.
03 · Decide
You get a clear verdict and what to fix first. Then we build it right, or walk away.
What you get
One report. A verdict, the patterns, and what to fix first.
Groundwork report
Platform Risk
Workflow automation vendor, support tickets to billing
1 failing3 at risk2 passing6 dimensions
- Incident historystrong evidenceFail
- Data handlinglimited evidenceRisk
- Compliance postureinferredRisk
- Vendor lock-inlimited evidenceRisk
- Security documentationstrong evidencePass
- Uptime recordstrong evidencePass
Fix first
Get the vendor's incident reports before you sign. Ask how billing errors get caught and reversed.
Evidence, not vibes.
Every finding shows where it came from. Claims don't count until something backs them up.
Vendor claim
limited evidence"Enterprise-grade security."
Independent verification
strong evidencePublished security report, checked against the vendor's trust page.
Services
Start free. Go deeper when it's worth it.
Free Groundwork scan
A quick readiness check on your idea. Free.
- Price
- Free
- Duration
- A few minutes
Fit call
We look at your case together. You leave with a next step.
- Price
- Free
- Duration
- 30 min
- Most popular
AI Readiness Audit
Full Groundwork report plus my expert review. UX, governance, and a fix-first plan.
- Price
- [TODO: price]
- Duration
- [TODO: e.g. 1 to 2 weeks]
Fractional AI UX Lead
Ongoing. I lead AI UX and product strategy inside your team.
- Price
- Monthly retainer
- Duration
- Ongoing
The honest part
I'll tell you if AI is the wrong move.
Even when it's not what you want to hear. That's the job.
Who's behind it
Nelly Pourmehr
AI UX Lead · Founder
I've worked in UX for about ten years. I led UX for Hypatos, an AI document automation platform used by 500+ enterprise users.
I wrote The Interface Is Not the System, about why control, trust, and responsibility break when AI systems scale. Groundwork is that book turned into a tool.
- AuthorThe Interface Is Not the System
- PlatformLed UX for an AI platform with 500+ enterprise users
- EducationMIT Professional Education
- Experience~10 years in UX
FAQ
Questions, answered plainly.
What is Groundwork?
Do I need AI already?
How long does an audit take?
Do you certify compliance?
Do you work remote?
What languages do you work in?
Have an AI decision to make?
Start with what you have.