Training Completed. Capability Unproven.

A completed course is evidence that someone completed a course.

It is not yet evidence that they can perform.

That distinction sounds obvious, but many capability systems still rely heavily on attendance, completion rates, certifications and knowledge checks because they are easy to record. They tell you who showed up, who finished and sometimes who remembered the material well enough to answer a test.

They do not tell you what happens when the skill has to be produced live.

New workforce research makes that gap unusually visible.

Pluralsight's 2026 Tech Skills Report, released on 30 September 2026, surveyed 1,500 technology executives, learning and development leaders and practitioners in the United States, United Kingdom and Australia. Seventy-seven percent of tech executives said they had seen employees complete training or earn certifications and still be unable to apply those skills in day-to-day work. Ninety percent of tech executives and 84 percent of L&D leaders were concerned that current skill measurement could be gamed or manipulated.

The study is about technology skills, not negotiation. But the measurement problem is highly transferable: participation is not the same as demonstrated capability.

A second report released the same day points in the same direction from the hiring side. Western Governors University's 2026 Workforce Decoded surveyed 3,128 US hiring professionals. Sixty percent said AI had made candidates' real skills harder to evaluate. Critical thinking was the most frequently cited skills gap, followed by effective AI use and verbal communication.

These are different surveys with different populations and should not be combined as if they measured one thing. Together, however, they reinforce a useful question for professional learning:

How do you know that someone can actually do what the training was supposed to build?

Negotiation exposes the problem immediately

Negotiation is a useful test case because the difference between knowing and doing becomes obvious very quickly.

A buyer may be able to define BATNA.

They may understand anchoring.

They may know why concessions should be conditional.

They may pass a multiple-choice test on all three.

Then the supplier says:

"Eight percent is the minimum. If you cannot accept that, we need to discuss whether we can continue supplying you next year."

Now the capability has to be produced.

Does the buyer challenge the claim?

Do they ask for evidence?

Do they separate the price demand from the continuity threat?

Do they reveal urgency?

Do they trade something valuable too early?

Do they hold the silence after asking a difficult question?

Do they adapt when the prepared script stops fitting?

None of those behaviours can be inferred reliably from course completion.

That is why negotiation should be treated as a performance skill rather than only a knowledge domain.

A knowledge test answers a different question

Knowledge assessment still matters.

If a buyer cannot explain what a BATNA is, they are unlikely to use the concept well. If they do not understand conditional trading, realistic practice alone will not magically supply the framework.

The mistake is not testing knowledge.

The mistake is treating a knowledge test as proof of execution.

A quiz can ask whether a concession should normally be exchanged for something in return.

A realistic conversation can show whether the buyer actually does that when a supplier keeps pushing, time is short and giving two percent would make the tension disappear.

Those are different measurements.

One tests recall and understanding.

The other tests behaviour under conditions that make the behaviour difficult.

Results alone are not enough either

There is another tempting shortcut: use the commercial outcome as the capability measure.

That also creates problems.

A buyer can achieve a large saving because the market was favourable, several suppliers were competing aggressively and the incumbent had priced high.

Another buyer can still face a cost increase after negotiating extremely well in a sole-source situation with no immediate alternative.

The result matters, but it does not isolate the negotiator's capability.

That is why our separate article on measuring negotiation performance distinguishes deal context, commercial outcome and decisions the buyer could actually control.

Capability evidence asks a related but different question:

Can this person repeatedly make commercially coherent decisions when the conversation becomes difficult?

What should negotiation capability evidence show?

Useful analysis does not need to reward a fixed bag of tactics.

The goal is not to count how often someone uses silence or whether they always anchor first.

Good negotiation depends on context.

The analysis should instead look for observable decision quality.

Did the buyer clarify before moving?

Did they test unsupported claims?

Did they protect information that did not need to be revealed?

Did they distinguish a real constraint from pressure language?

Did they trade rather than concede?

Did they maintain structure after escalation?

Did they recognise new information that genuinely justified a change of position?

Did they close the conversation with commitments clear?

These behaviours are observable because they appear in the conversation itself.

The important part is that they are interpreted in context rather than converted into a simplistic checklist.

The practice needs resistance

A practice scenario becomes weak if the simulated counterpart makes success too easy.

Someone cannot demonstrate the ability to handle pushback if the counterpart accepts the first reasonable argument.

They cannot demonstrate information discipline if the counterpart volunteers everything.

They cannot demonstrate composure under pressure if the conversation never creates pressure.

This is why realism matters in role-play. As we discuss in what makes AI role-play practice real, the counterpart needs credible constraints, limited information and believable resistance. Natural-sounding speech is not enough.

A scenario should create the conditions in which the target behaviour becomes necessary.

If the skill is holding a position after escalation, the scenario has to escalate.

If the skill is questioning a supplier cost claim, the supplier needs to make a claim that can be tested.

If the skill is trading rather than conceding, the counterpart needs to ask for movement.

One performance is not the person

There is an important safeguard.

A realistic simulation gives better evidence than a course-completion badge, but one conversation still should not become a permanent judgment of the employee.

People have off days. Scenarios vary in difficulty. Some weaknesses appear only in certain situations.

For capability building, the more useful pattern is repeated evidence across several targeted situations.

Does the buyer consistently lose structure when continuity risk is introduced?

Do they repeatedly reveal their target too early?

Do they improve after feedback?

Does the behaviour hold when the scenario changes?

That turns observation into development rather than ranking.

It also creates a much more useful management view. Instead of "92 percent of the team completed negotiation training," a leader can see which behaviours are stable, which pressure moments still create difficulty and whether practice is changing them.

Completion metrics are convenient. Capability evidence is useful.

Training teams need completion data. Compliance functions may need it. Administrators need to know who participated.

But completion should not carry more meaning than it deserves.

The recent Pluralsight findings are useful because they put a large number behind something many managers already recognise: people can finish training and still be unable to use the skill in real work.

Negotiation makes the distinction visible in minutes.

The buyer either produces the behaviour when the pressure arrives, or they do not yet produce it reliably.

That is not a failure of training.

It is the point where training needs to become practice, observation, feedback and another attempt.

A completed course can tell you that learning activity happened.

A realistic performance can start to tell you whether capability exists.

Sources

• Pluralsight (30 September 2026). 2026 Tech Skills Report.
• Western Governors University (30 September 2026). 2026 Workforce Decoded: 60% of employers say AI has made real skills harder to evaluate.