The Other Side Has AI Too

Artificial intelligence is supposed to reduce work.

That is the promise behind much of the current excitement. AI will automate routine tasks, eliminate administrative burden, accelerate analysis, and allow fewer people to accomplish more.

All of that may be true.

But it is only half of the equation.

AI does not just make existing work easier. It makes entirely new work cheap enough to create.

That distinction matters because most work does not exist in isolation. A complaint creates an investigation. A request creates a response. An RFP creates a proposal. A proposal creates a review. A contract question creates a legal analysis. An escalation creates a meeting.

When AI removes the friction required to initiate those activities, the resulting work does not disappear. It moves to someone else.

A few months ago, I pulled into an airport parking garage. The signs clearly stated that the first floor was reserved for valet parking. No problem. I continued toward the second floor to self-park.

On the ramp between the first and second floors, I found an open space. I interpreted the ramp as neither the first floor nor the valet section. I parked, caught my flight, and returned the next day.

There was a $45 ticket on my windshield.

The local police department and court system apparently interpreted the location differently.

Before AI, I probably would have paid the ticket.

Not because I agreed with it, but because fighting a $45 parking ticket was not worth the effort. I would need to determine how to appeal it, find the right contact, explain the facts, write a coherent argument, attach evidence, track the response, and follow up.

The cost was not really $45. The cost was $45 versus an hour or more of my time.

AI changed the economics of that decision.

I took a photo of the ticket, a photo of where I parked, and gave ChatGPT a two-sentence explanation. Within moments, I had a clear, professional appeal. It explained my interpretation of the signage, documented the ambiguity, laid out the argument for dismissal, provided a subject line, and identified how the request should be framed.

The entire process took about two minutes.

I sent the appeal. A month passed with no response. I followed up and said that, given the lack of a reply, I assumed the ticket had been dismissed.

They confirmed that it had.

From my perspective, AI worked exactly as advertised. It saved time, reduced friction, and helped me reach a fair outcome.

From the perspective of the police department and court system, it created another piece of work.

Someone had to receive the appeal. Someone had to review the ticket, consider the photographs, interpret the parking rules, decide whether the citation should stand, update the record, and respond to the follow-up.

My productivity gain became their incoming workload.

That is the part of the AI story we are underestimating.

AI Removes the Friction That Prevented Work

People have always had reasons to dispute a fee, question a policy, challenge an invoice, appeal a decision, or ask for clarification.

Most of those actions never happened because they required effort.

The customer did not have time to read the contract. The homeowner did not understand the HOA bylaws. The executive did not know how to interpret the technical report. The buyer did not want to write the RFP. The employee did not have time to turn a half-formed idea into a polished proposal.

That friction acted as a filter.

Sometimes the filter prevented legitimate questions. Sometimes it prevented unnecessary ones. Usually it did both.

AI removes much of that friction.

Have a dispute with your homeowners association? You no longer need to read a hundred pages of articles, declarations, and bylaws. Upload the documents, explain the issue, and ask AI to identify the relevant provisions and draft a formal letter.

Do not understand a line on an electrician's invoice? Upload it. Within minutes, AI can explain the terminology, compare the work to common practice, and prepare five questions to ask the electrician.

Receive a technical root cause analysis after a service outage? AI can translate the report into plain language, identify inconsistencies, and generate a list of follow-up questions. You can also upload the service-level agreement, terms of service, and master services agreement to understand whether you may be owed credits and draft the request.

Need a new provider? AI can create a detailed RFP in moments.

Every one of these examples is useful to the person initiating the request.

Every one of them creates work for the person receiving it.

This is not necessarily good or bad. It cuts both ways.

AI gives individuals and smaller organizations access to capabilities that once required lawyers, analysts, consultants, procurement teams, or technical specialists. That can improve accountability and produce better decisions.

It can also flood organizations with requests that would never have existed before.

The efficiency gained by one party may be paid for through increased volume on the other side.

The RFP Arms Race

Consider the traditional enterprise procurement process.

A buyer needs a new provider. In the past, the buyer might have spoken with a handful of vendors, developed a short requirements document, requested proposals, and worked through the details during the sales process.

Now the buyer can ask an AI system to produce a twenty-page RFP covering technical requirements, security controls, service levels, legal provisions, implementation methodology, governance, reporting, compliance, support, transition planning, and pricing.

The buyer did not suddenly develop a twenty-page set of requirements. The buyer acquired the ability to produce one.

The document is sent to five vendors.

Each vendor now has to respond.

Fortunately, each vendor also has AI.

The vendor uploads the RFP, internal product documentation, prior proposals, security statements, architecture diagrams, and contractual language. The AI system generates a fifty-page response.

Five vendors return 250 pages of material.

The buyer cannot reasonably read it all.

Fortunately, the buyer has AI.

The responses are uploaded, summarized, scored, compared, and ranked. The system generates a list of gaps, risks, follow-up questions, and negotiation points.

The vendors receive another round of questions.

They use AI to respond.

No one in this process is necessarily behaving irrationally. Each participant is using technology to become more efficient, thorough, and responsive.

Yet the total system may be producing far more work than it did before.

The RFP is longer. The responses are longer. The review is more detailed. The follow-up questions are more numerous. The decision process becomes more elaborate because every participant can create more material at nearly zero marginal cost.

This is not automation eliminating work.

It is automation multiplying it.

The Internal Version May Be Worse

The same dynamic will occur inside companies.

Employees will use AI to create more memos, proposals, analyses, business cases, operating plans, and executive updates.

Ideas that once remained as brief conversations will become polished documents. Questions that once died in someone's inbox will become formal requests. A passing thought can become a twelve-slide presentation before lunch.

The cost of contribution has collapsed.

That sounds positive, and in many cases it will be. More people can communicate clearly.

Employees who struggle with writing can express strong ideas more effectively. Teams can analyze issues with greater depth. Leaders can access more perspectives.

But polished output is not the same as valuable thinking.

AI makes it easy to produce a document that looks complete before the underlying idea is complete.

That creates a new kind of internal AI slop.

The problem is not only low-quality public content. It is the volume of professional-looking material entering the operating system of the company.

Every memo asks someone to read it. Every proposal asks someone to evaluate it. Every analysis asks someone to debate it. Every escalation asks someone to resolve it. Every update creates the expectation of acknowledgment.

Organizations may soon discover that they have dramatically increased the capacity to produce information without increasing the capacity to make decisions.

The bottleneck moves.

Writing becomes cheap. Judgment remains expensive.

Analysis becomes abundant. Attention remains scarce.

Ideas become easy to package. Execution remains difficult.

The Other Side Has AI Too

In the first chapter of Harry Potter and the Half-Blood Prince, the British Prime Minister becomes frustrated after hearing about the havoc caused by Voldemort and the Death Eaters.

The Prime Minister insists that the Minister for Magic should be able to solve the problem. After all, he has magic.

The response is simple: "The trouble is, the other side can do magic too".

That is a useful way to think about AI.

Your customers have AI.

Your vendors have AI.

Your employees have AI.

Your regulators have AI.

Your competitors have AI.

The person appealing the decision has AI, and the organization reviewing the appeal has AI.

The procurement team has AI, and every bidder has AI.

The plaintiff has AI, and the defendant has AI.

The employee making the case has AI, and the executive evaluating it has AI.

AI makes each individual participant more capable. It does not necessarily make the system more efficient.

In fact, when every participant uses AI to generate more complete arguments, more documentation, more questions, and more responses, the overall system may become significantly busier.

This is why predictions that AI will lead directly to widespread unemployment may prove too simplistic.

AI will eliminate certain tasks. It may eliminate some roles. It will allow organizations to operate with fewer people in specific areas.

But it will also unlock enormous amounts of previously suppressed demand.

More disputes will be filed because they are easier to file.

More analysis will be requested because it is easier to produce.

More providers will be evaluated because procurement is easier to initiate.

More proposals will be submitted because they are easier to write.

More contracts will be scrutinized because expertise is easier to access.

More ideas will be circulated because they are easier to package.

The volume of work may expand faster than the efficiency used to process it.

We should be careful not to confuse the automation of a task with the elimination of the activity surrounding that task.

The Cost Will Not Be Shared Equally

In consumer markets, businesses will usually bear most of the cost.

A customer can generate a detailed complaint, appeal, or service request in minutes. The business must receive it, interpret it, investigate it, and respond.

The customer may spend two minutes. The company may spend hours.

As more consumers gain access to better arguments, deeper contract analysis, and professional communication, businesses will face higher volumes of more sophisticated requests.

Some of those requests will be legitimate and overdue.

Some will be unnecessary.

Most companies will not be able to distinguish between the two without reviewing them.

In business-to-business relationships, the cost will be more evenly distributed.

Both sides will create more documents. Both sides will demand more detail. Both sides will ask more questions. Both sides will use AI to defend their position.

The result may be an administrative arms race in which every company becomes more productive individually while the ecosystem becomes less efficient collectively.

New Friction Will Emerge

Friction is usually treated as something to eliminate.

But some friction serves a purpose.

It forces people to decide whether an issue is important enough to pursue. It limits low-value requests. It encourages prioritization. It prevents every possible question from becoming an active workstream.

As AI removes that friction, organizations will likely introduce new forms of it.

Businesses may limit the number of submissions a customer can make. They may restrict the length of requests. They may require structured forms instead of open-ended letters. They may prioritize requests based on customer value, severity, or contractual obligation.

Some may charge fees for certain reviews, appeals, or proposal processes.

Others will deploy stronger filtering systems to determine which requests deserve human attention.

Many will simply accept higher engagement volume as a permanent operating cost.

Cost itself may become the most important constraint. AI usage is inexpensive today, but not free. If generating, submitting, evaluating, and responding to every possible interaction carries a real economic cost, that cost may provide some discipline.

What seems less likely is a world in which autonomous AI systems simply negotiate with each other and remove humans from the process.

The reason is not technical. It is institutional.

The most important decisions involve judgment, accountability, trust, liability, and competing interests. Humans will remain involved because someone has to own the outcome.

AI may prepare the argument, evaluate the response, and recommend the next step.

A person will still decide whether to send it.

Efficiency Is Not the Same as Less Work

The first era of AI adoption is focused on individual productivity.

Can this employee write faster? Can this team analyze more data? Can this company automate more tasks? Can this customer solve a problem without hiring an expert?

Those are reasonable questions.

The next era will be about system-level effects.

What happens when every participant can create more work for every other participant?

What happens when the cost of asking approaches zero, but the cost of deciding remains high?

What happens when every document can be expanded, every argument strengthened, every issue escalated, and every interaction prolonged?

AI may make us more efficient at producing work while making the world less efficient at absorbing it.

That does not mean the technology is bad. My parking ticket probably should have been dismissed, and AI helped make that happen.

It means the benefits will not appear evenly.

One person's saved hour may become another organization's new case.

One team's faster RFP may become five vendors' longer responses.

One employee's polished proposal may become an executive's additional decision.

The other side has AI too.

And that may create far more work than it eliminates.