The Tale of Two AI DRaGs

Carl Burch
Jun 7
6 min read

Updated: Jun 8

Typically, when you hear the term drag, you naturally think of something that’s holding you back from accomplishing your tasks and objectives. However, if we’re talking about Artificial Intelligence (AI), the term has two distinct meanings: Operational AI Drag and Productivity AI Drag.

Operational AI Drag

Operational AI Drag refers to outsourcing work to AI. Outsourcing work to AI involves strategically assigning tasks, workflows, and certain decisions to an AI system. Unlike traditional outsourcing, which shifts work to external people or firms, AI outsourcing replaces human labor with algorithmic processing, allowing businesses to scale without adding headcount. The important issue for any company is to know whether outsourcing diminishes or enhances its distinctive competency.

The Importance of Distinctive Competency

A distinctive competence is a capability that sets a company apart from its competitors, such as producing higher-quality products, offering better customer service, or being more innovative.

As a manager, you must identify what makes your company unique and protect it by building barriers that competitors cannot easily copy. That is how a business stays competitive and survives.

A core principle of strategic management is that companies should not outsource their distinctive competencies. That remains true in the age of AI. Whether AI outsourcing weakens an advantage depends on what is outsourced and who owns the underlying data model.

If a company uses generic public AI tools for core proprietary work, it risks weakening the advantage that sets it apart. But when AI supports non-core activities or is built into proprietary models trained on the company’s own data, it can strengthen and protect that advantage.

In this context, Operational AI Drag stands for:

Drafting (using AI to generate emails, outlines, and first drafts)
Research (using AI to summarize articles, trends, and competitive data)
Analysis (using AI to outline pros and cons and build frameworks)
Grunt work (using AI for tasks like formatting tables and checklists)

These are the non-core tasks best suited for AI. When delegated well, AI can complete them faster, improve consistency, and lower costs. That frees management to focus on the core activities that drive business success and create competitive advantage.

Protect Your Advantage

To prevent AI from eroding your competitive advantage, use the simple strategic filter below to understand whether to outsource an activity or build it internally.

Productivity AI Drag

Now, let’s turn to Productivity AI Drag. This occurs when a business uses AI tools poorly, reducing productivity and ROI. Productivity Drag commonly appears in the following forms:

Prompt Wrestling: This happens when getting the right result from AI takes more time than doing the task yourself. It often shows up in email drafting, where repeated edits and refinements outweigh the time savings AI was supposed to provide.
The Overproduction of "AI Workslop": “AI Workslop,” a term coined by researchers at Stanford Social Media Lab and BetterUp Labs, refers to polished AI-generated output—such as lengthy reports, emails, or code—that looks finished but lacks substance or context. Passing this kind of low-value work downstream wastes time for colleagues who must interpret, fix, or redo it.
Desynchronized Workflows: Desynchronized workflows occur when different team members use AI tools at varying speeds, formats, and quality levels, breaking the collaborative link between the departments. As an example, let’s say your marketing team uses AI to instantly generate 50 product pitches, but if the design team still operates on a manual, two-week sprint cycle, the workflow desynchronizes. This results in bottlenecks and severe communication breakdowns.
Early-Career "AI Slowdown": Early-Career "AI Slowdown" occurs when junior employees use AI to skip the hard, messy process of learning core skills, which severely stunts their long-term professional growth.

Ways to Mitigate AI Drag

If you’re looking for ways to mitigate AI drag, the first thing you must do is to stop treating it [AI] as a “magic shortcut” and start treating it as a core infrastructure with standardized rules, quality controls, and proper system integration. The image below shows what is meant by moving from a “shortcut” mindset to a structured operation framework.

Below, we examine each of the four “AI drags” described above and suggest ways to reduce their impact.

1. Mitigate Prompt Wrestling

Replace open-ended chat boxes with structured input forms. Employees have to fill out specific fields (e.g., "Target Audience," "Core Product Feature"), and the backend automatically wraps that data in an engineered, highly optimized system prompt.
Mandate that prompts include 3 to 5 flawless human-written examples directly into the system prompt. LLMs recognize patterns faster than they follow abstract descriptions, eliminating stylistic guessing games.
Stop giving the AI complex prompts. Prompt wrestling usually happens when an employee tries to get the AI to do a bunch of stuff at once, which it can’t do (e.g., "Read this report, find the errors, summarize it, and write an email about it"). Make the instructions “simple stupid.”

2. Mitigate the Overproduction of "AI Workslop"

Introduce strict corporate standards. If an employee uses AI to expand a minor thought into a 10-page report, they must use AI to compress it back down into a 3 to 5-bullet-point executive summary before sending it to a human recipient.
Mandate that the core structure, hypothesis, or data analysis must be executed by a human first.
“Workslop” spreads when accountability disappears. To counter this, companies should implement documentation rules that require employees to prove they actively engaged with the AI's output.

3. Mitigate Desynchronized Workflows

Set explicit limits on how much work an AI-empowered team can pass forward at one time. A content team might use AI to draft 20 articles a day, but the operational gate must limit submissions to 3 articles a day to match the manual editing team's capacity.
Align the definition of a "completed task" across departments. If a developer uses AI to write code in 10 minutes, that task is not considered "done" until the automated QA pipeline and human code review are completed.
Ensure all departmental AI tools feed into a single, unified database or CRM (such as Salesforce Einstein or Microsoft Copilot) rather than living in unlinked point solutions.

4. Mitigate Early-Career "AI Slowdown."

Require new hires or junior employees to complete critical tasks manually for their first 60 to 90 days. This may seem inefficient, but routine work is often how employees learn their roles. It helps them build a strong understanding of core processes, logic, and internal workflows before they begin automating those tasks with AI.
Train juniors as peer reviewers. Give them flawed AI outputs and then have them identify hallucinations, logic gaps, and context errors. This builds the editing and critical-thinking skills they need to work effectively with AI.
Pair junior employees with senior experts in mandatory co-piloting sessions. Instead of having juniors work alone with an LLM, seniors can model real-time critical thinking and show them how to question, verify, and tailor AI-generated work.

Final Word

Today’s “AI battle” is less about which company or country leads the race and more about balancing short-term efficiency with the long-term development of human talent. While tech giants dominate the headlines, small and medium-sized businesses face a more immediate challenge: choosing between instant output and investing in the future skills of their workforce.

When companies use AI to automate entry-level work, they may gain short-term efficiency but lose the training ground where junior employees learn the basics of the business. In the process, they can weaken a distinctive competency that once gave them a competitive edge. That is the paradox of rushing to adopt AI.

Early evidence points to junior corporate roles being more vulnerable than many blue-collar jobs: the Oliver Wyman Forum and NYSE 2026 CFO Agenda survey found that 64% of CFOs expect their finance functions to shift away from junior roles over the next three years.

Here’s what few people are talking about...

Aviation offers a clear warning about overreliance on technology. Because autopilot handles routine flying, pilots spend less time flying manually. The danger emerges when systems fail and control suddenly returns to the pilot at the precise moment expert skill is needed, yet that skill may have eroded from lack of practice. In aviation, this is known as “automation dependency.” (LinkedIn)

That is where the real “AI battle” will be fought: between those who treat AI as the answer to everything and those who understand that outsourcing critical thinking to algorithms creates workers who can complete tasks but cannot judge results.

About Carl Burch

Carl Burch holds an MBA, CMA, CIA, FCCA, and is a QuickBooks ProAdvisor. He is also the co-founder of BURCH Business Services (BBS) located in Boston, MA.

You can contact Carl at carl.burch@burchbusinesservices.com, or for more information on how BBS can help you improve your business operations, visit www.burchbusinesservices.com