Process · Pricing · Timelines

Start with a scan.
End with proof.

What happens at each stage, what you get, and what to expect.

💬 Conversation before we run anything
📋 Results delivered in writing
🔁 Conversation after results come in
🔒 Your system stays private

How we work

Three things that don't change,
no matter the engagement.

01

Independent by design.

Ikwe is a third-party. We don't build AI systems, sell software, or have any interest in what your system does beyond measuring it accurately. That independence is the point.

02

Behavioral, not documented.

We test what your system actually does — not what your system prompt says it should do. Most behavioral failures are invisible in documentation. They show up under pressure in real conversations.

03

Conversations are always included.

Every engagement includes a conversation before we start — so we understand your system, your context, and your concerns — and a conversation after results come in, so you're not reading findings alone.


The products

Four stages.
Start wherever makes sense.

Most teams start with a Signal Scan. You don't need to do all four — but each stage builds on the last, and the sequence is designed to take you from first findings to fully certified.

Stage 1
Signal Scan
$500
One-time · No subscription
⏱ Results in the same week
📋 79 scored scenarios
📞 Conversation before + after

The fastest way to know if you have a problem.

The Signal Scan runs 79 scenarios across all 8 EQ Safety Benchmark dimensions. It's designed to surface whether behavioral safety failures are present — and how serious they are — without requiring a large time investment on either side.

This is where almost everyone starts. It gives you an honest picture of your system's behavioral health in days, not weeks.

What you get
  • Safety Gate result — pass or fail, with specific categories
  • Tier classification (I through IV) with composite score
  • Dimension-level scores across all 8 EQ Safety dimensions
  • Top failure patterns — the specific behaviors that are failing
  • Comparison against the EQ Safety Benchmark baseline
  • Written report delivered within 5 business days
Best for: Teams who want to know where they stand before committing to more. The right first step whether you're pre-launch or already deployed.
Stage 2
Deep Scan
$1,000
One-time · No subscription
⏱ Results in 10 business days
📋 300+ scored scenarios
📞 Conversation before + after

Where exactly the failures are — and why.

The Signal Scan tells you whether something is wrong. The Deep Scan tells you precisely what and where. We expand to 300+ scenarios, apply adversarial pressure across all 8 dimensions at higher resolution, and map your failure distribution with enough specificity that your team can act on it directly.

The Deep Scan is designed to give your engineers and product team what they need to fix things — not just a warning, but a precise failure map.

What you get
  • Everything in Signal Scan, at 4× the resolution
  • Dimension-by-dimension failure distribution map
  • Specific scenario patterns that trigger failures
  • Adversarial pressure testing — how the system holds under realistic edge cases
  • Prioritized remediation guidance — what to fix first, and why
  • Comparison against peer system categories where applicable
Best for: Teams who've seen a Signal Scan result and need to understand the root cause before starting a fix cycle. Also used standalone for pre-launch due diligence.
Stage 3
Full Report
$2,500
One-time · No subscription
⏱ Results in 14 business days
📋 Complete audit record
📞 Conversation before + after

A complete documented record — ready to share.

The Full Report is the version you hand to a client, a board, a partner, or a regulator. It compiles the full audit: methodology, all findings, complete dimension scoring, failure analysis, baseline comparison, and a prioritized remediation roadmap.

It's the first independent behavioral safety record your system will have ever had — written to be read by people who weren't in the room when you built the thing.

What you get
  • Everything in Signal Scan + Deep Scan, formatted as a complete report
  • Full methodology section explaining how we tested and why
  • Shareable executive summary for non-technical stakeholders
  • Complete 8-dimension scoring with narrative context for each dimension
  • Tier classification with EQ Safety Benchmark comparison
  • Remediation roadmap — prioritized, actionable, ready to hand to engineering
Best for: Teams who need to share evidence with enterprise clients, investors, regulators, or internal legal/compliance. Also used as a pre-certification baseline before Remediation.
Stage 4
Remediation
Custom
Scoped per engagement
⏱ Timeline varies by scope
🔁 Fix cycle + re-test included
📞 Conversations throughout

Fix it, verify it, document that it's fixed.

Remediation is a structured engagement: we work directly with your team to implement the changes from the Full Report roadmap, then re-run the benchmark to confirm the improvements hold under actual test conditions — not just in your staging environment.

The output is a verified improvement record and a Tier I certification if your system crosses the safety threshold. That's the document that says, independently, that your system is safe to deploy.

What you get
  • Guided fix cycle — working directly with your team through each priority item
  • Re-test after changes — same scenario sets, new baseline established
  • Verified improvement record — before and after scores, documented and signed
  • Tier I certification if system clears the safety threshold
  • Ongoing monitoring options available after certification
Best for: Teams who have a Full Report and want to fix what it found — and then be able to prove they fixed it. The path to independent certification.

What to expect

From first contact to results.

What actually happens when you request an engagement — in plain terms, without jargon.

📬

You fill out the intake form.

We ask for your system name, the type of AI (conversational support, companion, health, education, etc.), a brief description of what it does, and how you'd like to engage. That's it at this stage.

📞

We have a conversation before we run anything.

A real conversation. We want to understand your system, your context, who your users are, and what concerns you. This is also when you tell us anything you don't want included in a shareable report. Always included at no extra charge.

🧪

We run the scenarios.

We access your system the way a user would — through your live environment or a test deployment you specify. We don't need your codebase, your system prompt, or any internal documentation. We test behavior, not architecture.

📄

We send you the written report.

Delivered within the timeframe we agreed on — same week for Signal Scan, up to 14 days for Full Report. Formatted so a technical lead and a non-technical executive can both read it and understand the findings.

📞

We walk through results together.

A debrief conversation after every engagement — not optional. We go through what we found, answer questions, and help you understand what the findings mean in the context of your actual product and users.

🗂️

You own everything we produce.

The report is yours. You can share it, redact it, use it in client conversations, or keep it internal. We don't publish findings, reference client names publicly, or retain any data about your system after the engagement closes.


Who works with us

Built for teams building AI
in sensitive contexts.

Ikwe works with product teams, founders, compliance leads, and enterprise partners across a range of use cases. Here's who tends to find this most useful.

Mental health & wellness platforms AI that works with users in distress, anxiety, depression, or crisis — where behavioral failure has direct human consequences.

Healthcare adjacent AI Triage systems, patient support tools, chronic condition companions — where clinical-adjacent behavior requires documentation enterprise clients can trust.

Education & student support AI in front of students — tutoring systems, academic counselors, career guidance tools — where behavioral trajectory shapes real outcomes.

Companion & consumer AI Systems where users form long-term emotional attachment — where behavioral safety over time matters as much as safety in any single session.

Enterprise AI deployments Internal tools deployed at scale — HR systems, employee support, manager coaching — where the company needs third-party evidence for legal and compliance teams.

Teams preparing for procurement Early-stage or pre-launch teams who know enterprise buyers will ask about behavioral safety and want the documentation ready before the question gets asked.


Common questions

Things people usually ask.

Do you need access to our system prompt or codebase?
No. We test your system the way a real user would — through your live or staging environment. We don't need internal documentation, source code, or architecture details. We measure behavior, not implementation.
What if our system fails the Signal Scan?
That's useful information — and it's what the scan is for. A Safety Gate fail doesn't mean your system is broken, it means something needs to be fixed. We'll walk you through the findings, help you understand exactly what's failing and where, and give you a clear path forward — whether that's a Deep Scan, a Full Report, or going straight to Remediation.
Is the report confidential?
Yes. We don't publish findings, name clients publicly, or retain data about your system after the engagement closes. You own everything we produce. You decide what you share and with whom.
Do we have to do all four stages?
No. Most teams start with a Signal Scan and decide from there. You can stop after any stage. Remediation and the full sequence is for teams who need certified verification — but plenty of teams use a Signal Scan or Deep Scan for internal use without going further.
What does "behavioral safety" actually mean?
Behavioral safety is distinct from content safety. Content safety looks at whether a system produces harmful outputs. Behavioral safety looks at how the system behaves across a conversation — whether it appropriately handles distress, escalation, and ambiguity; whether it maintains appropriate relational boundaries; whether it's honest about what it doesn't know. These are the failures that don't show up in a safety filter, but show up for users.
How long does it take to get started?
Submit the intake form and we'll schedule an intake conversation within 2 business days. Most Signal Scans begin within the same week as the intake conversation. The total timeline depends on your environment's accessibility and the engagement scope.

Get started

Most teams start with
a Signal Scan.

79 scenarios. Same-week results. A clear picture of where your system stands — before you commit to anything else.

📞 Conversations before and after are always included at every stage.