The Augmented Human: How to Work Better with AI Without Outsourcing Your Own Judgment
By Pedro Sonda, dooop
Working better with AI means using a method: first deciding if the task is one AI performs well, providing sufficient context, requesting verifiable results, actually verifying them, and assuming the final decision. The "augmented human" is the one who combines AI's speed with their own judgment. Without a method, AI increases both workload and errors; with a method, it improves quality.
Updated in October 2026.
What Research Shows About Working with AI
In an experiment with 758 BCG consultants, conducted by researchers from Harvard, Wharton, MIT, and Warwick, those who used AI on tasks within AI’s strengths completed 12.2% more tasks, 25.1% faster, and with quality more than 40% higher. For tasks outside that scope, they had a 19 percentage point lower chance of being correct (Dell'Acqua et al., 2023).
The researchers identified two patterns among those who used AI well:
- Centaur: divide the work, delegating to AI the parts it is strong at and keeping the rest.
- Cyborg: integrate AI at every step, in a constant back-and-forth between person and system.
Interpretation: both patterns work. What does not work is accepting AI’s output without knowing if the task was within or outside AI’s capabilities.
dooop Framework: The Augmented Work Cycle
| Step | What to Do | Control Question |
|---|---|---|
| 1. Delimit | Decide if the task, or part of it, is suitable for AI | Does AI usually get this type of task right? Is the cost of an error high? |
| 2. Contextualize | Provide objective, audience, constraints, examples, and quality criteria | Would a new employee understand the request? |
| 3. Request Verifiable Output | Ask for sources, assumptions, calculations, and uncertainties | Can I check where each statement came from? |
| 4. Verify | Check facts, numbers, logic, and fit to context | What would happen if this were wrong? |
| 5. Decide and Record | Take responsibility for the decision and record what was done with AI, when relevant | Would I sign off on this result? |
Step 4 in Detail: Proportional Verification
Not everything requires the same level of verification. A draft internal email calls for a quick read; a number presented to the board requires source verification. A Microsoft Research study with 319 knowledge workers found that higher trust in AI was associated with less critical thinking (Lee et al., 2025). Verification must be a deliberate habit precisely because the natural tendency is to relax when AI gets several things right in a row.
The 4 Most Common Bad Habits
| Bad Habit | Consequence | Alternative |
|---|---|---|
| Copying and pasting without reading | Errors and fabricated information reach the recipient | Read everything before using |
| Requesting without context | Generic results and rework | Provide objective, audience, and criteria |
| Using AI for tasks requiring personal judgment | Fragile decisions with no clear ownership | Use AI to prepare the decision, not to make it |
| Entering sensitive data into any tool | Risk of exposing company and client data | Follow usage policy and use approved tools |
In a study with over 48,000 people in 47 countries, 56% reported having made work errors because of AI (KPMG and University of Melbourne, 2025). Most of these errors are avoidable with steps 1 and 4.
For Executives: AI in Decision Preparation
Executives can use AI to enhance decision preparation without delegating decisions:
- Map alternatives: request options the team has not considered.
- Test arguments: ask for the strongest argument against the preferred decision.
- Summarize evidence: consolidate long reports, verifying cited numbers at the source.
- Simulate questions: anticipate what the board, clients, or regulators will ask.
AI’s role is to improve the quality of preparation. The decision and responsibility remain with the executive. See AI Does Not Replace Judgment: What This Means for Leaders.
What to Do Now
- Adopt the 5-step cycle on a recurring task for one week and compare the results.
- Define with your team which types of deliverables require source verification.
- Share AI errors and successes in short rituals; this is how the team learns where the boundary lies.
To build teams using this method, learn about dooop’s AI literacy at dooop.com.br/letramento-em-ia.
Frequently Asked Questions
Does using AI make people less capable? It can, if the person stops thinking about what AI produces. Used with verification and judgment, AI tends to expand capacity in suitable tasks.
Should I disclose when I used AI in a deliverable? It depends on company policy and the type of deliverable. For client documents, technical reports, and relevant decisions, recording the use and review is good practice.
Also Read
- Which Human Capabilities Become More Important with AI?
- Generative AI, Agents, and Automation: What’s the Difference?
About dooop
dooop is a consultancy based in Caxias do Sul (RS) that prepares leaders and organizations to transform artificial intelligence into business capability through strategy, governance, and training.