How do I stay in control?
AI is making decisions.
Who is responsible?
When AI starts making decisions in your work or business, someone has to be responsible for the results. These tools and services help you stay in control — and prove it when it matters.
Two Pillars. One Framework.
Governance
The how of running agents responsibly
Governance defines the operational layer — who oversees agent decisions, how outputs are reviewed, what triggers human escalation, and how the system stays aligned with its actual purpose as context shifts. It is the ongoing practice, not the one-time document.
Oversight frameworks
Human-in-the-loop protocols
Audit cadence and review cycles
Escalation and exception handling
Agent scope boundaries
Compliance
The what of proving it to others
Compliance is the documentation and verification layer — the audit trail, the liability map, the regulatory assessment. It is what you show a CFO, legal team, or regulator when they ask how decisions were made and who was responsible for them.
Liability mapping and responsibility matrices
Regulatory exposure assessment
Audit trail documentation
Deployment sign-off frameworks
Board and legal-ready reporting
What AI Cannot Be Trusted to Do on Its Own
Someone raised a concern with us recently: can AI be trusted by communities that have historically been misrepresented by the systems built to serve them?
It is the right question. The honest answer is: not without human oversight.
AI systems trained on human-generated data inherit human biases — racial, cultural, economic, and linguistic. This is documented, not theoretical. The communities most likely to be harmed by biased AI outputs are often the communities least represented in the data those systems were trained on.
This is not a reason to refuse AI entirely. It is a reason to stay in the process — to verify outputs before they become decisions, to name who is accountable, and to retain the right to override what the agent produces.
These are not technical safeguards. They are human ones. And they are exactly what Pertinent's tools are built around.
Documented Risk
Racial and cultural bias in training data
AI models reflect the data they were trained on. That data overrepresents certain demographics, languages, and cultural contexts — and underrepresents others. Outputs may systematically disadvantage communities not well-represented in training data.
Documented Risk
Linguistic bias
AI performs better in dominant languages and registers. People who communicate in minority languages, dialects, or non-standard forms may receive lower quality outputs — affecting everything from hiring tools to healthcare communication to educational assessment.
Systemic Risk
Economic bias in what problems AI solves
AI tools are predominantly built for problems that affect those who can afford to fund their development. The needs of under-resourced communities, nonprofits, and public institutions are systematically underserved by the market.
Accountability Risk
Invisible decision-making
When AI makes or influences a decision — a loan approval, a hiring screen, a content filter — the person affected often has no way to know, question, or contest it. Accountability requires transparency about when and how AI is involved.
The Human Oversight Response
01
Review before it becomes a decision. The Output Auditor exists because AI outputs require human judgment before they act on anyone's behalf. This matters most when the people affected have least recourse.
02
Name who is accountable. When AI is involved in a decision that affects a person, someone must be accountable for that outcome. The Agent Accountability Checklist makes that accountability explicit.
03
Retain the right to override. Checkpoint 03 of the checklist is Override — your right to stop, redirect, or reject what the agent produced. This is not a technical setting. It is a human decision that must be made consciously.
04
Be transparent with the people affected. If AI is involved in something that affects your students, clients, or community members, they have a right to know. Transparency is not a legal technicality — it is the foundation of trust.
Pertinent does not resolve the structural problems of AI bias. No single tool can. What these tools do is keep the human in the process — specifically the human who knows their community, their context, and their responsibilities — at every point where an AI agent might otherwise act without oversight.
Why This Matters Now
When generating content and code becomes nearly free, the scarce resource shifts. Verification, judgment, and trust become the bottleneck. AI agents are now making decisions inside real organisations — scheduling, drafting, routing, executing — faster than the governance frameworks designed to contain them.
The result is a verification gap most organisations have not mapped. Agents executing faithfully against the wrong brief. Outputs that go unreviewed. Decisions with no audit trail. Liability that sits with the human who pressed deploy — whether they understood the exposure or not.
"Verification is the new scarcity. When intelligence is cheap, the capacity
to verify, judge, and take accountability for agent outputs becomes the scarce
and valuable resource. Scale without verification is a liability that compounds."
Every productive workflow in the agentic economy follows one structure: Human Intent → Agent Execution → Human Verification. The agent executes in the middle. The human owns the beginning and the end. Pertinent builds the tools for both.
Where the Risk Lives
High Exposure
Unaudited Outputs
AI outputs acted upon without human review. No record of what the agent produced, when, or why. When something goes wrong, there is no trail.
High Exposure
Undefined Liability
No clear answer to who is responsible when an agent makes a bad decision. The organisation, the vendor, the employee who configured it?
Medium Exposure
Scope Creep
Agents doing more than they were designed to do because no one defined the boundary. Access, authority, and action scope left open-ended.
Medium Exposure
Regulatory Blind Spots
GDPR, HIPAA, SOC 2, and emerging AI-specific regulation applying to agent behaviour in ways the deploying organisation has not assessed.
High Exposure
Unverified Agent Identity
No standardised way to prove what an agent is authorised to do, who it represents, or what constraints govern it. When something goes wrong, there is no verifiable record of what the agent was permitted to do in the first place.
High Exposure
Unintended Delegation
A single approval triggering multi-step workflows nobody intended. Users authorise an agent without understanding the full scope of what they have enabled — and failures go unreported with no clear path to diagnosis.
The Pertinent Compliance Stack
Each tool targets a specific compliance gap. Use them independently or as a framework.
00
Readiness Scorecard
60 seconds. Ten questions. An instant score that tells you what percentage of your work is agent-ready today — and which band you are in. No email required. The first step before any other tool.
Live
01
Agent Stack Audit
The deeper diagnostic. Assess readiness before deployment — which processes are agent-ready, which carry risk, and what ROI to expect. The entry point for every compliance conversation.
Live
02
Agent Delegation Auditor
Pre-flight check before an agent runs. Paste your instruction — receive a plain-language breakdown of what you are authorising, what access it implies, what could go wrong at each step, and a risk rating. Before you hand over control.
Coming Soon
03
Output Auditor
Review AI outputs before they become decisions. Flags accuracy issues, hallucination risk, gaps, and bias. Creates a reviewable record of human oversight — the audit trail that proves the human was in the loop.
Live
04
A policy targets a compliance gap. Five questions that produce a plain-language AI use policy in your own words — for educators, nonprofits, small businesses, independent professionals, and healthcare workers. The document that defines what you and your organisation commit to before an agent runs. Free, no sign-up required.
Live
05
Five plain-language checkpoints for anyone using AI agents in work that serves students, clients, or communities. Entry, Instruction, Override, Audit, Exit. The document that keeps the human present at every point where an agent might otherwise act without oversight. Free PDF download.
Live
06
Agent Scope Definer
Define exactly what an agent is allowed to do, access, and decide. Produces a boundary specification that prevents scope creep, stops unintended multi-step workflows, and creates a governance baseline.
In Development
07
Liability Map Generator
Map who is accountable for each agent decision in your workflow. Produces a clear responsibility matrix — the document a CFO, COO, or legal team needs before signing off on autonomous deployment.
In Development
08
Agent Identity Brief
Document what your agent is, what it is authorised to do, and what it has actually done — in a format any auditor, regulator, or board member can read. KYA (Know Your Agent) compliance before standards are formalised.
In Development
09
Compliance Brief Generator
Input your industry and agent use case. Receive a plain-language assessment of applicable regulations and the minimum governance steps required before autonomous deployment. The document your board needs.
In Development
Consulting Engagements
The tools give you a starting point. The engagements deliver the full framework — built for your specific workflows, team, and regulatory context.
Compliance Architecture Review
A structured assessment of your current or planned agent deployments against a governance framework. Produces a liability map, a risk register, and a prioritised remediation plan. Delivered in two sessions.
Enquire →
Agent Governance Framework
A bespoke governance document for your organisation — defining agent scope, human oversight requirements, audit procedures, and escalation protocols. The document your board or legal team needs to proceed with confidence.
Enquire →
Start with the free tools.
Scale with the framework.
The Agent Stack Audit takes three minutes and tells you exactly where your compliance exposure sits right now.