Facial recognition shows how AI can boost productivity and harm rights at the same time. Here’s how to use AI at work without repeating Mobile Fortify’s mistakes.
AI facial recognition is here. The question is how we use it.
A single scan of a phone in the hands of an ICE officer can pull from more than 200 million images and multiple federal databases in seconds. Name, date of birth, immigration status, outstanding warrants — all surfaced faster than a human could flip through a single file.
That’s an example of AI technology doing what it does best: compressing hours of manual work into a few seconds. But it’s also an example of what happens when productivity is treated as the only metric that matters.
This post is part of our AI & Technology series, where the usual focus is on using AI to streamline work, boost productivity, and make better decisions. Here, we’re looking at a harder edge of the same story: what AI does when it’s pointed at people in public spaces, not just spreadsheets, and why the same principles that make AI useful at work have to include guardrails, ethics, and respect for rights.
What happened in Chicago — and why it matters beyond immigration
A 23‑year‑old U.S. citizen, Jesús Gutiérrez, was walking home from the gym in Chicago when an unmarked SUV rolled up. ICE officers got out, questioned him, handcuffed him, and put him in the car. He had no physical ID on him. Instead of taking time to verify his story through traditional checks, they took a photo of his face.
On the other end of that photo was Mobile Fortify, a facial recognition app used by ICE and CBP. It runs a face scan against a huge stack of government databases and images — including immigration records, FBI data, and state warrant systems. A few moments later the agents said, essentially: the system says he’s good. They dropped him off after about an hour.
Here’s why this story is bigger than one encounter:
- The technology worked fast and produced an answer. From a productivity standpoint, that’s exactly what many teams want from AI.
- The stop itself appears to have been based on how he looked and where he was, not on specific evidence of a crime or immigration issue.
- The same system that can quickly confirm someone’s identity can also misidentify, overreach, or override physical documentation, according to members of Congress and civil liberties groups.
Most companies building or adopting AI tools will never be in the business of street-level immigration enforcement. But the logic is the same: once you adopt AI to make faster decisions, you have to decide what trade‑offs you’re willing to accept and what constraints you’re willing to enforce.
How Mobile Fortify works — and what it says about AI at scale
The core idea behind Mobile Fortify is straightforward: give field agents a high‑speed, AI‑driven way to identify people using facial recognition on a smartphone.
According to internal documents and reporting:
- Agents use a work phone to scan a person’s face.
- The app compares that image against a database of roughly 200 million images.
- It pulls data from multiple government systems, including immigration and law enforcement databases.
- It returns a profile: name, date of birth, immigration status, and other flags that might matter for enforcement.
From a pure technology and productivity perspective, this is very familiar:
- It’s a workflow optimizer: compressing data lookups, cross‑checks, and paperwork into one interface.
- It’s AI‑powered pattern matching: similar to what teams use to match customers to records, flag fraud, or prioritize support tickets.
- It’s mobile‑first: designed so the decision happens “in the field,” not back at a desk.
That’s exactly the pattern we see across AI in work and productivity:
- Feed AI large, messy datasets
- Add fast pattern recognition (facial recognition, text analysis, anomaly detection)
- Put it in the hands of people who make time‑sensitive decisions
The catch is simple: speed and scale amplify both good and bad decisions. If the initial decision to stop someone is biased, AI makes that bias faster and more efficient.
The productivity trap: when “efficiency” collides with rights
Here’s the thing about AI for work and productivity: everyone wants faster. Fewer clicks. Less manual searching. More automation. That’s the same instinct that drove Mobile Fortify.
But in the Chicago case and others like it, we can see three traps that apply just as much to a small business rolling out AI tools as to a federal agency with a facial recognition app.
1. Treating AI outputs as “definitive” truth
Internal briefings to lawmakers suggest that some ICE officials see a Mobile Fortify match as a “definitive” determination of status — strong enough to ignore other evidence, even a birth certificate.
That mindset shows up in business all the time:
- A risk model flags a customer as high‑risk, so the account gets closed without human review.
- A CV‑screening AI filters out qualified candidates based on pattern‑matching past hires.
- An internal AI assistant summarizes a contract incorrectly, and no one checks the source.
Productivity tip that actually protects you: AI should be a first pass, not a final verdict. The higher the stakes (freedom, finances, jobs, safety), the more you need human review, not less.
2. Scaling existing bias faster
Gutiérrez is of Mexican descent. He was stopped in a neighborhood where immigration enforcement is common. Critics now refer to similar stops based on race, language, or location as “Kavanaugh stops” — shorthand for profiling focused on people who “look” like they might be undocumented.
AI doesn’t create that bias out of nowhere. It amplifies what’s already there:
- If your sales routing algorithm assumes big cities are “better” leads, rural customers get ignored.
- If your support triage tool prioritizes “standard English” messages, non‑native speakers wait longer.
- If your internal monitoring flags only certain teams for strict review, they operate under constant suspicion.
When you plug AI into a biased process, you don’t just get more efficiency. You get more efficient unfairness.
3. Optimizing for the wrong metric
Mobile Fortify optimizes for speed of identification, not for fairness of who gets scanned, or accuracy across all demographics, or respect for civil rights.
In work contexts, the same thing happens:
- Targeting only “time saved” leads to broken experiences for customers you didn’t optimize for.
- Automating decisions purely for “throughput” risks regulatory trouble and brand damage.
- Chasing “more data” without guardrails leads to privacy incidents that erase trust you can’t easily rebuild.
This matters because AI is now baked into how we work: from calendar assistants to CRM automations to code copilots. Choosing the wrong optimization target is how teams get efficient at the exact things they shouldn’t be doing.
What ethical AI looks like in practice (not just policy decks)
If you’re building or deploying AI — whether you’re a startup founder, a team lead, or a solo operator — you can take concrete lessons from how Mobile Fortify is being used and criticized.
Ethical AI isn’t vague. It’s operational. Here’s what that looks like.
1. Make “human in the loop” non‑negotiable for high‑impact decisions
Any AI system that can affect someone’s rights, job, or access to services should assist, not replace, human judgment.
In practice:
- Treat AI outputs as recommendations, not commands.
- Require manual review for high‑risk outcomes: account closures, legal decisions, hiring rejections, bans, terminations.
- Log when humans overrule AI — those edge cases are where you learn the most.
The opposite approach — treating the model as unquestionably right — is how you end up in the same mindset as “the app says this person is an alien, ignore their documents.”
2. Limit where and how the tech is used
The DHS documents acknowledge that Mobile Fortify can be used on U.S. citizens and lawful permanent residents, even though it was framed internally as a tool to find people who can be removed from the country.
Well‑run AI deployments have clear boundaries:
- Define who can use an AI tool and under what conditions.
- Explicitly state where the tool must not be used.
- Build in technical friction — for example, requiring a case ID or justification before a high‑impact decision can be executed.
For a workplace AI system, this could mean:
- Your AI analytics tool can’t see HR medical data.
- Your customer‑data bot can’t export raw PII without a second approval.
- Your internal AI assistant can’t access legal documents marked as privileged.
3. Collect less data than you think you “could”
Mobile Fortify’s power doesn’t just come from clever algorithms. It comes from the volume and sensitivity of the data it taps: immigration files, FBI records, warrant databases, and more.
In business settings, most teams wildly over‑collect:
- Full chat transcripts when you only need issue summaries
- Full customer browsing history when you only need the last few interactions
- Biometric data or location history when simpler signals would do
The more sensitive the AI’s input, the stronger your privacy and security requirements have to be. For 99% of productivity and workflow cases, you don’t need anything like the depth of Mobile Fortify’s data to get value.
4. Measure error — and who pays the price
Facial recognition systems are known to perform unevenly across races, genders, and age groups. Even if the overall error rate looks low, that doesn’t tell you who’s getting misidentified.
Do the same audit on your own AI:
- Where does it fail most often?
- Which customers or employees bear the cost of those failures?
- When it’s wrong, how hard is it to fix the record or appeal the decision?
Ethical AI isn’t about never being wrong. It’s about knowing how you’re wrong and designing recovery paths that are actually humane and fast.
Using AI at work without repeating Mobile Fortify’s mistakes
Most readers of this series care about AI, technology, work, and productivity because you want to get more done with less stress and fewer manual tasks. You’re not building surveillance tools. But the same underlying questions still apply:
- Who does this AI impact?
- What happens when it’s wrong?
- Which incentives are we baking into the system?
Here’s a practical checklist you can use before deploying any new AI tool in your workflow:
- Define the real goal. Is it fewer support tickets, faster onboarding, better creative ideas? Avoid vague targets like “more AI.”
- Map the risks. Could this hurt reputations, finances, mental health, or rights? Higher risk ≠ no AI, but it does mean more safeguards.
- Set decision boundaries. Write down which decisions AI is allowed to make alone and which always require people.
- Document data flows. Know exactly what data the system touches, where it goes, and how it’s stored.
- Plan for appeals and overrides. Make it easy for people — customers, employees, partners — to challenge AI‑driven decisions.
- Review regularly. AI isn’t “set and forget.” As your data and use cases change, your guardrails need to adjust with them.
Working smarter with AI isn’t just about squeezing out inefficiencies. It’s about designing systems you’d be comfortable being subject to yourself. If you’d be uneasy with strangers scanning your face on the street and trusting an app over your own documents, that’s a signal for how carefully we should be rolling out AI in boardrooms, back offices, and browser tabs.
Where this leaves us
Facial recognition in public spaces is a clear, high‑stakes example of AI’s double edge. The same features that make AI powerful productivity tech — speed, scale, pattern recognition — can also make it oppressive when pointed in the wrong direction, with the wrong rules.
If you’re building or adopting AI at work, the lesson from Mobile Fortify is blunt: efficiency without ethics is a liability, not an advantage. The teams that will actually win this decade aren’t the ones that cram AI into everything. They’re the ones that pair AI with clear boundaries, accountability, and respect for the people on the other end of the decision.
As you design your next AI‑powered workflow, ask a simple question: Would this still feel fair if I were the one being analyzed, scored, or flagged? If the honest answer is no, you don’t need more AI. You need a better plan.