VAMP Is Measuring the Smoke. What If We Could See the Fire Starting?
VAMP Is Measuring the Smoke. What If We Could See the Fire Starting? Why Merchant Intelligence and Product Intelligence May Become the L...
There is a question being asked across almost every industry right now:
Will AI replace my job?
Payments isn't immune.
As artificial intelligence moves deeper into merchant onboarding, underwriting, fraud, risk and compliance, it's reasonable for underwriters to wonder what their role looks like five years from now.
Yes.
It will do many of those things.
In fact, some of them are already happening.
But I don't believe AI is going to eliminate the need for skilled human underwriters anytime soon.
I believe something much more interesting is about to happen.
AI is going to eliminate much of the work that prevents underwriters from actually underwriting.
And that's a very different conversation.
Ask an executive and you might hear:
Technically, that's correct.
Now sit beside an underwriter for a day.
Watch what actually happens.
Then, somewhere inside all of that work:
Make a risk decision.
Which one of those activities actually required the experience and judgment of an accomplished underwriter?
That's the question we should be asking.
Underwriting has always required investigation.
That's not the problem.
The problem is how much of the investigation remains manual.
Highly trained risk professionals spend enormous amounts of time finding information before they can begin interpreting it.
There is an important distinction between those activities.
Investigation asks:
What information can I find?
Intelligence asks:
What does the information mean?
Underwriting asks:
What should we do about it?
Those should not be the same job.
Yet our legacy infrastructure forces the underwriter to perform all three.
One AI merchant-risk provider, Ballerine, currently describes the problem this way: most underwriters spend about 70% of their time on manual reviews. Its platform positions AI agents as a way to transform fragmented merchant data into risk profiles so underwriters can concentrate on decisions rather than collection.
Whether the exact percentage inside your organization is 70%, 60%, or 40% isn't really the point.
Measure it yourself.
Ask:
That's the metric that matters.
Because every hour a senior underwriter spends manually searching websites, collecting evidence, comparing fields, reviewing routine documents, or writing repetitive summaries is an hour you're paying for expertise but consuming labor.
This is where AI becomes transformative.
Modern systems can ingest enormous amounts of structured and unstructured information.
They can crawl websites.
Those capabilities align remarkably well with the most labor-intensive portion of underwriting.
The objective shouldn't necessarily be:
AI → Decision
It should increasingly be:
AI → Intelligence → Human Judgment → Decision
That distinction matters.
Imagine opening an underwriting case and instead of beginning with an application and 17 attachments, you receive an intelligence package.
Business Identity
Verified.
Ownership
Validated.
Licensing
Identified and checked.
Sanctions / Watchlists
Screened.
Adverse Media
Relevant findings summarized.
Website
Analyzed.
Business Model
Classified.
Products
Identified.
MCC
Recommended with supporting rationale.
Restricted Activity
Flagged.
Marketing Claims
Analyzed.
Documents
Extracted and evaluated.
Product Compliance
Exceptions identified.
Policy Requirements
Applied.
Evidence
Captured.
Changes
Tracked.
And instead of 200 pieces of raw information, the underwriter receives:
Three material issues requiring professional judgment.
Now the underwriter is doing what you actually hired them to do.
Underwrite.
This isn't theoretical.
A new generation of merchant-risk platforms is already approaching underwriting this way.
Coris describes its platform as AI merchant-risk infrastructure spanning onboarding, underwriting, monitoring and other risk workflows. Its AI assistants can research merchants, analyze patterns and summarize cases, while its platform centralizes merchant intelligence for risk teams.
Ballerine uses AI agents for merchant underwriting and risk management, including website analysis, adverse media, MCC classification, policy logic and evidence-backed audit trails. Importantly, Ballerine explicitly frames the goal as turning underwriters and risk professionals from data collectors into decision makers.
And at Qredible, this is exactly why we built Q-Trust™.
Q-Trust applies AI-enabled automation and Product Intelligence™ to enhanced due diligence, website and product discovery, product classification, documentation, licensing, marketing claims, applicable compliance requirements and continuous monitoring. Qredible says Q-Trust automates more than 80% of regulated merchant EDD while preserving a digitized evidence chain across merchants, products and compliance artifacts.
Different platforms have different strengths.
But collectively they demonstrate something important:
The underwriting technology stack is changing.
There's an understandable fear that automation diminishes the value of the human.
I think the opposite happens when it's implemented correctly.
Consider two underwriters.
Underwriter A
Processes 10 cases.
Spends hours collecting information.
Manually searches websites.
Reads routine documents.
Copies data.
Creates screenshots.
Writes summaries.
Performs repetitive checks.
Finally makes 10 decisions.
Underwriter B
Has technology continuously collecting, structuring, analyzing and documenting the underlying information.
Processes 40 cases.
Spends the majority of the day evaluating exceptions.
Investigates unusual business models.
Evaluates conflicting evidence.
Challenges AI findings.
Applies institutional risk appetite.
Communicates with sales and compliance.
Handles difficult judgment calls.
Makes 40 informed decisions.
Which underwriter is more valuable?
I'd argue Underwriter B.
Not because AI replaced their expertise.
Because AI amplified it.
Anyone who has spent meaningful time underwriting merchants knows that risk doesn't always fit neatly inside a rule.
Businesses are messy.
Applications are imperfect.
Information conflicts.
Business models evolve.
There are exceptions.
There are nuances.
A merchant can look risky and be perfectly legitimate.
Another can look pristine and be hiding something.
Sometimes the most important signal isn't an individual data point.
It's the relationship between several of them.
Experienced underwriters develop intuition from thousands of cases.
They understand context.
They ask the question that wasn't in the checklist.
They recognize when something simply doesn't add up.
They understand the institution's actual appetite, not merely its written policy.
They know when to escalate.
And they understand consequences.
That judgment remains enormously valuable.
We should be equally clear about this.
AI isn't infallible.
Models can hallucinate.
Data can be incomplete.
Sources can be wrong.
Context can be misunderstood.
Automated classifications can be incorrect.
Rules can conflict.
Merchant structures can be unusual.
Novel risks can emerge that weren't anticipated by the system.
And an AI system doesn't personally own the risk appetite of your institution.
That's why blindly replacing human judgment with an opaque algorithm isn't the future I would advocate.
The objective isn't human out of the loop.
It's:
Human at the right point in the loop.
This should become a fundamental design principle for modern underwriting.
If the information is verifiable, verify it.
If data can be extracted, extract it.
If a website can be scanned, scan it.
If products can be identified, identify them.
If policy can be applied automatically, apply it.
If evidence can be captured automatically, capture it.
If nothing changed, don't make an analyst review it again.
If the merchant clearly meets established policy, don't manufacture unnecessary friction.
But when the evidence conflicts...
When risk signals interact...
When the business model is unusual...
When the merchant falls outside established policy...
When an exception could be commercially valuable...
When the consequences are material...
Give it to a human.
That's where expertise belongs.
Legacy underwriting asks humans to inspect everything so they can discover what matters.
Modern underwriting should use technology to inspect everything and ask humans to evaluate what matters.
That's a profound difference.
The operating model changes from:
COLLECT → SEARCH → READ → VERIFY → COMPARE → DOCUMENT → ANALYZE → DECIDE
to:
AI COLLECTS
↓
AI STRUCTURES
↓
AI ANALYZES
↓
AI APPLIES POLICY
↓
AI CAPTURES EVIDENCE
↓
AI SURFACES EXCEPTIONS
↓
HUMAN EVALUATES
↓
HUMAN DECIDES
That is not replacing underwriting.
That is modernizing underwriting.
There's another benefit that doesn't get enough attention.
When underwriting decisions and supporting evidence become structured data, organizations can learn from them.
Why was this merchant approved?
What signals mattered?
Which exception was granted?
What happened afterward?
Which merchants subsequently generated losses?
Which underwriting indicators were predictive?
Which were noise?
Which policy produced unnecessary declines?
Which exceptions became successful merchants?
Where do underwriters disagree?
Suddenly underwriting itself becomes measurable intelligence.
The organization can improve its models.
Improve policy.
Improve training.
Improve consistency.
And improve future decisions.
AI isn't simply helping the underwriter.
The underwriter can help make the intelligence system smarter.
This is one of the most important parts of the transformation.
A fast decision isn't necessarily a good decision.
And an automated decision isn't necessarily a defensible decision.
For risk and compliance teams, we need to know:
What was evaluated?
What was discovered?
Which evidence supported the finding?
Which policy was applied?
What exceptions existed?
Who made the decision?
What changed afterward?
Can we recreate the decision six months later?
Modern platforms are increasingly designed around this principle. Ballerine, for example, emphasizes explainable assessments supported by structured data and audit trails. Coris describes human-in-the-loop escalation and audit trails within its AI-agent architecture.
Q-Trust similarly maintains timestamped evidence across merchant and product compliance activity, designed to support ongoing monitoring and future examinations.
That's enormously important.
Because the future of underwriting isn't simply faster decisions.
It's more defensible decisions.
This brings us directly back to the previous articles in this series.
When a policy changes, what happens to yesterday's underwriting?
In the manual model, much of the intelligence used to make the original decision disappeared into:
PDFs.
Screenshots.
Analyst notes.
Spreadsheets.
Emails.
Case files.
Or someone's memory.
So the organization has to investigate again.
But imagine the original underwriting process created structured intelligence.
Merchant.
Products.
Documents.
Evidence.
Findings.
Risk factors.
Policy relationships.
Decision.
All digitized.
Now when policy changes, the institution isn't necessarily beginning another investigation from zero.
It can reevaluate existing intelligence against the new requirement.
That's where AI becomes bigger than underwriting automation.
It becomes institutional memory.
This may be the biggest mindset change.
Today we often think about underwriting as a process.
Application enters.
Underwriter reviews.
Decision occurs.
Case closes.
I believe underwriting should increasingly be viewed as an intelligence-creation event.
Every merchant investigated should make the institution smarter.
Every product identified should become structured intelligence.
Every document analyzed should become usable data.
Every risk signal should be retained.
Every decision should preserve its rationale.
Every subsequent change should enrich the profile.
Then your portfolio becomes more than a collection of merchant accounts.
It becomes a living intelligence asset.
That's enormously powerful.
This is where the implications become even more interesting.
Why do processors avoid certain industries?
Sometimes because the financial risk is genuinely unacceptable.
Sometimes because regulatory or reputational risk falls outside appetite.
But sometimes—as discussed in the previous articles—it's simply because the merchant is too expensive to understand and monitor manually.
Enhanced due diligence costs money.
Product reviews cost money.
Website reviews cost money.
Ongoing monitoring costs money.
Policy changes create re-audits.
Therefore:
Complex Merchant = Expensive Merchant = High Risk Merchant
But what happens when AI dramatically reduces the cost of understanding complexity?
The economics change.
Suddenly some categories that were uneconomical to underwrite manually may become commercially viable.
And that creates something much bigger than cost savings.
It creates market opportunity.
Imagine two processors competing for the same regulated merchant.
Processor A says:
Processor B says:
Who wins?
Now multiply that across thousands of merchants.
Better underwriting intelligence doesn't simply reduce risk.
It can increase acceptance precision.
Reduce unnecessary declines.
Improve merchant experience.
Accelerate onboarding.
Expand supportable markets.
Strengthen sponsor-bank confidence.
And help sales pursue opportunities competitors cannot economically evaluate.
That makes AI underwriting infrastructure something more than a compliance investment.
It becomes growth infrastructure.
I believe the role changes.
Less:
Researcher.
Data collector.
Document processor.
Website inspector.
Spreadsheet manager.
More:
Risk analyst.
Exception specialist.
Policy interpreter.
Commercial advisor.
Intelligence validator.
Decision maker.
Portfolio strategist.
That's an upgrade.
And the best underwriters will likely become substantially more productive and strategically valuable.
It's that your competitors give their underwriters AI and you don't.
Imagine two organizations five years from now.
One has underwriters manually searching websites, reviewing PDFs, copying information between systems and documenting findings.
The other has AI continuously discovering, validating, structuring and monitoring merchant intelligence while experienced professionals concentrate on exceptions and decisions.
Which organization:
Boards merchants faster?
Needs fewer incremental hires to grow?
Finds more hidden risk?
Responds faster to policy changes?
Creates better audit evidence?
Supports more complex merchants?
Provides a better sales experience?
Maintains greater consistency?
Can manage a larger portfolio?
The answer isn't difficult.
And that's why I believe the conversation about AI replacing underwriters is asking the wrong question.
Not:
Ask:
Not:
Ask:
Not:
Ask:
Not:
Ask:
“How do we build the most intelligent underwriting organization in the market?”
That's the strategic opportunity.
AI will change underwriting.
Dramatically.
Pretending otherwise would be foolish.
But I don't believe the future of merchant underwriting is a giant algorithm approving and declining businesses while humans disappear from the equation.
I believe the future is far more powerful.
Machines doing machine work.
Search.
Extraction.
Classification.
Comparison.
Monitoring.
Calculation.
Pattern recognition.
Evidence collection.
And:
Humans doing human work.
Context.
Judgment.
Challenge.
Exception management.
Risk appetite.
Commercial reasoning.
Accountability.
Decision.
Put those capabilities together and you don't get fewer underwriters.
You get better underwriting.
Faster.
More consistent.
More scalable.
More explainable.
More defensible.
And capable of supporting forms of commerce that today's manual economics make extraordinarily difficult.
So no.
And the payment companies that figure that out first may discover that their underwriting department isn't merely a function they can make more efficient.
It's a competitive advantage they can make exponentially more powerful.
Qredible is redefining merchant underwriting through Merchant Risk Intelligence (MRI)—a product-first approach that continuously analyzes what businesses sell, how they market those products, and the evidence required to support compliant payment acceptance. By moving beyond static industry classifications, Qredible helps banks, payment processors, ISOs, and sponsor banks make faster, more informed, and more defensible underwriting decisions while reducing manual effort and strengthening ongoing portfolio oversight. Learn more about Qredible's product-first automated compliance management platform for regulated industries →
