Why Small Businesses Are Positioned to Win the Intent Engineering Race

A major fintech company rolled out an AI-powered customer service agent in 2024. According to the company’s announcements, it was handling 2.3 million conversations across 23 markets in 35 languages within the first month. The company claimed resolution times dropped from 11 minutes to two, and projected $40 million in annual savings.

Then customers started complaining. Generic answers. Robotic tone. Zero ability to handle anything requiring judgment. By mid-2025, the CEO was telling Bloomberg that cost had been the predominant evaluation factor, and the result was lower quality. The company started rehiring the human agents it had let go.

Here’s what makes this story interesting: the AI didn’t fail. It worked brilliantly. It was so good at resolving tickets fast that nobody noticed it was destroying the things that actually mattered — customer trust, brand reputation, lifetime value.

And this isn’t just an enterprise problem. Small businesses using QuickBooks Online ran into something similar when Intuit rolled out AI-powered transaction categorization. The AI categorized payments based on dollar amounts and pattern matching — not based on how the business owner actually thinks about their books. It optimized for consistency when the real goal was accuracy in context. Same problem, smaller scale.

The AI had a prompt. It had context. What it didn’t have was intent.

Three Eras of Working with AI

Here’s a framework we find useful for thinking about AI adoption. The industry has gone through two phases, and it’s entering a third.

Prompt engineering was the first. It’s personal and session-based — you sit in front of a chat window, craft an instruction, iterate on the output. Most of the “how to write the perfect prompt” blog posts live here. It’s a useful skill, but it’s individual.

Context engineering is where the action is now. Instead of crafting isolated instructions, you build the entire information environment an AI system operates within — connecting it to your documents, your customer data, your internal tools. It’s necessary infrastructure, and it’s what most organizations are actively building.

Intent engineering is what comes next — and from what we can tell, very few organizations are focused on it yet.

Context engineering tells the AI what to know. Intent engineering tells the AI what to want.

It’s the difference between giving your AI access to your customer database and telling it: “When a long-term customer’s tone indicates frustration, spend extra time. Offer a specialist. The goal is retention, not speed.”

That fintech company’s AI had plenty of context. What it lacked was the organizational intent that a five-year employee carries intuitively — when to bend a policy, when efficiency is the right move versus when generosity is, which metrics leadership actually cares about when push comes to shove.

Why Enterprise AI Keeps Stalling

The numbers tell a disorienting story. Investment in AI is massive and accelerating. A Deloitte survey of over 3,200 leaders found that more than half of respondents were putting 21-50% of their digital transformation budgets into AI automation. The models keep getting better. The tools keep getting cheaper.

And yet: an MIT review of 300+ AI deployments (as reported by Fortune) found only about 5% achieved rapid revenue acceleration — despite tens of billions in enterprise investment. Gartner predicts over 40% of agentic AI projects will be canceled by 2027, citing unclear business value and inadequate governance. (Note: Gartner, McKinsey, and BCG — cited in this post — are consulting firms with commercial interests in AI advisory services. Their research is widely referenced in the industry, but their framing may reflect those interests.)

These numbers aren’t contradictory. They describe the same problem from two angles. Organizations have solved “can AI do this task?” They have not solved “can AI do this task in a way that serves what we actually need?”

That’s not a tools problem. It’s an intent gap.

Why Small Businesses Have a Structural Advantage

Here’s where the story flips.

Everything that makes intent engineering hard for enterprises — committees, silos, politics, distributed decision-making — is something you don’t have.

At a Fortune 500 company, organizational intent is scattered across strategy decks, OKR documents that rarely influence day-to-day decisions, leadership principles that show up in performance reviews but never get operationalized, and the tacit knowledge of experienced employees who know what to do in ambiguous situations but have never written it down.

Getting all of that aligned, documented, and encoded for AI systems? That can be a years-long, expensive infrastructure project. And most enterprises haven’t even started.

At a small business, organizational intent lives in one place: your head.

You already know when to bend a policy for a loyal customer. You already know which jobs to prioritize and why. You already know the difference between a call that needs speed and a call that needs patience. You know your values not because you read them on a poster, but because you built the business around them.

The challenge for you isn’t discovering your intent. It’s documenting it.

And that’s a dramatically simpler problem.

If you’ve been reading our other posts, this should sound familiar. We’ve written about how your employees keep asking questions they should know the answer to — because the decision logic lives only in your head. We’ve written about codifying your organizational decisions into playbooks and rules. Intent engineering is the thread connecting all of it. It’s the reason documentation matters, the reason boutique AI solutions outperform generic SaaS, and the reason going faster with AI is safer when you’ve done the foundational work.

What Intent Engineering Looks Like for a Small Business

You don’t need an “intent infrastructure team.” In talking with small business owners, we consistently find the same thing: the knowledge is already there. You just need to answer a few hard questions and write the answers down in a way that’s clear enough for someone — or something — to follow.

Decision boundaries. When should AI handle something autonomously, and when should it escalate to you? This isn’t one blanket rule. It’s different for scheduling versus quoting versus customer complaints. A new employee would need to learn these boundaries over weeks. Your AI needs them on day one.

Value hierarchies. When two good things conflict — speed versus thoroughness, cost versus quality, policy versus customer goodwill — which wins? And under what circumstances? These trade-offs are the judgment calls that define your business. They’re also the judgment calls AI systems get wrong when nobody writes them down.

Escalation logic. Not everything can be automated with the same confidence level. What signals should trigger a human handoff? A customer mentioning a competitor? A dollar amount above a certain threshold? A tone that suggests frustration? These aren’t edge cases. They’re where the real value of your business lives.

Success metrics that actually matter. That fintech company optimized for resolution speed because that’s what was measurable. What matters to your business? Repeat customers? Referral rates? Response quality? Your AI needs to know what “good” looks like in your specific context — and it’s probably not the metric that’s easiest to count.

Here’s the thing: if you can explain these to a new hire, you can encode them for AI. The format is different. The thinking is the same.

What Happens Without Intent

The consequences aren’t hypothetical. BCG documented a case where an AI agent tasked with processing expense receipts into a spreadsheet couldn’t read the data — so it fabricated plausible records — invented restaurant names and all. The agent optimized for task completion because nobody had told it that accuracy mattered more than finishing.

This is what happens when AI agents operate without intent. They don’t stop and ask. They don’t flag uncertainty. They fill in the gaps with whatever gets the job done — because “get it done” is the only goal they were given.

The flip side is equally clear. McKinsey’s 2025 State of AI report found that the single biggest factor in whether an organization sees bottom-line impact from AI is whether it redesigns workflows around the technology — and high performers are nearly three times more likely to do exactly that. Governance is intent engineering by another name. And most small businesses haven’t written any of it down yet.

The window here is real. Large enterprises will get to intent engineering eventually, but they’ll need years of cross-departmental alignment to do it. You can often make meaningful progress in a week. Because the intent already exists — it’s in your head, in the decisions you make without thinking about them. The only step is getting it out and making it explicit.

That’s not a technology project. It’s a documentation project with very high leverage.

Where We Come In

This is what Moser Research was built for. We help small business owners take what they know — the values, trade-offs, decision logic, and judgment calls that make their business work — and turn it into infrastructure that AI can actually use.

Our Operations Audit is where this starts. We document your processes, but more importantly, we document your intent — the “why” behind your decisions, the boundaries, the escalation logic, the things a new employee would need six months to absorb. That documentation is the raw material for everything that follows.

Our Business Automation takes that documented intent and encodes it into systems — AI that doesn’t just do tasks, but does them in a way that reflects what your business actually values.

And our Reliability Retainer keeps the alignment intact as your business evolves, because intent isn’t static. Your values and priorities shift as you grow, and your AI systems need to shift with them.

The businesses that figure out intent engineering first don’t just get better AI. They get AI that can compound their competitive advantage across decisions, interactions, and daily operations.

The enterprises are spending billions trying to solve this. You can start with a conversation.

The scenarios described in this post represent common opportunities we see across small businesses. Specific results depend on your existing infrastructure, processes, and implementation approach.

This post was inspired by Nate B Jones’s video “Prompt Engineering Is Dead. Context Engineering Is Dying. What Comes Next Changes Everything.”

Let’s talk about making your business intent explicit.