TL;DR. Sales training gets forgotten because the format is structurally optimized for knowledge transfer, not for behavior change. Brandon Hall Group and the Ebbinghaus forgetting curve have shown the same thing consistently for decades: 70 to 87 percent of delivered content is lost after 30 to 90 days, unless repeated application follows. The problem is not bad trainers. It is a missing architectural layer between the workshop and the customer conversation. This piece explains the 70/87 percent trap, why it destroys L&D budgets, and which layer a modern training system has to add to produce measurable behavior change in sales.

Key Takeaways

• 70 to 87 percent of classic training content is forgotten after 30 to 90 days (Brandon Hall Group, Ebbinghaus), regardless of trainer quality or method.
• Workshops deliver knowledge. Customer conversations demand behavior. Between the two sits a gap that no trainer and no slide deck can close.
• The missing layer is a practice-with-feedback loop. Until 2024 it was not scalable, because it required a coaching relationship per employee.
• AI coaching platforms like Sleak close this gap voice-natively, with scorecard-based feedback and endlessly repeatable practice sessions, GDPR-compliant and EU-hosted.
• Anyone who wants to scale sales coaching has to move from an event format to a system format. In 2026, that is no longer an aspiration. It is an operating decision.

Why sales training gets forgotten (and what the 70/87 percent trap means for L&D)

Every sales organization knows the scene. A two-day workshop, strong ratings at the end, motivated reps on Sunday evening, and three weeks later the team is back in its old pattern. Someone calls the trainer, because "the content did not stick." That is not a trainer problem. It is an architecture problem. And it has been predictable with unsettling precision since Hermann Ebbinghaus in 1885.

This piece explains why classic sales training is structurally condemned to be forgotten, what the 70/87 percent trap means for L&D budgets at DACH companies, and which architectural layer a modern training system has to add to actually change behavior in the customer conversation. It is written for L&D leaders, VPs of Sales, and people development leaders at mid-market and enterprise organizations who have to choose between repeating the workshop, extending the LMS, and building a real coaching architecture.

Why do sales reps forget what they learn in training?

Sales reps forget training content because human memory decays exponentially without repetition and application. Brandon Hall Group has measured this consistently for years: 30 days after a classic workshop, 70 percent of the content is lost, and after 90 days it is often 87 percent. The phenomenon is not specific to sales training. It is a basic principle of human learning. Hermann Ebbinghaus first described the underlying forgetting curve in 1885, and over more than a century of replication studies the shape of the curve has barely moved.

Three mechanisms drive the decay. The first is the absence of active application. A rep who heard on Tuesday how to counter price objections with value arguments has no practice loop between that and the real customer conversation on Wednesday. Without repetition in a realistic context, the content disappears from working memory long before it can turn into procedural skill.

The second is the format. Workshops are high-density knowledge blocks, optimized for maximum content in minimum time. That density is a delivery advantage and a retention disadvantage at the same time. The brain can process only a fraction of what an experienced trainer covers in eight hours, per session. The rest flows through, leaving an impression but no skill.

The third is context dependency. Content learned in the seminar room, at the whiteboard, stays bound to the seminar room in the rep's mind. The real discovery call with a skeptical CFO looks different. The CFO does not stick to the script, the call has its own dynamic, and the pressure is different. Knowledge that was never practiced in the context of its application does not reliably transfer across contexts.

What does the 70/87 percent trap mean for L&D budgets?

The 70/87 percent trap means that every euro spent on classic sales training produces, on average, 70 to 87 cents of loss, measured by learning retention after 30 to 90 days. At a typical mid-market sales training budget of 150,000 to 400,000 euros a year, that means a six-figure amount disappears into the forgetting curve. The number is not polemical. It is the direct consequence of Brandon Hall data and standard training economics at DACH companies.

The trap has one vicious property. It is invisible at the ROI level. A workshop gets a 4.5-star rating at the end, the trainer gets a follow-up engagement, and HR reports a 98 percent attendance rate up the chain. Nobody measures the gap between "the reps attended the workshop" and "the reps behave differently in the following quarter." The LMS dashboard cannot architecturally see that gap. It was never built for it. More on this structural weakness in the behavioral-change-at-scale argument and in the Completion Is Not Competence argument.

At DACH mid-market companies with 50 to 500 sales employees, the mechanics typically look like this. The training budget per rep runs between 1,500 and 4,000 euros a year, depending on industry and function. At 100 reps and an average of 2,500 euros, that comes to 250,000 euros. Apply the 70/87 percent rate, and 175,000 to 217,000 euros of that go into decay, if nothing happens between the workshop and the customer conversation. The remainder, roughly 30,000 to 75,000 euros, is the effective investment in behavior change. Per rep, the actually effective share of the training budget is therefore between 300 and 750 euros a year.

Most L&D teams know this. They are reluctant to talk about it, because the budget comes out of people development's past and gets defended in the logic of that past. The question they rarely ask openly is not whether the money disappears into the forgetting curve. It is how to change that without destroying the budget that justifies the L&D job in the first place.

How exactly does the forgetting curve work in sales training?

The Ebbinghaus forgetting curve shows an exponential decline in retention over time, with the steepest loss in the first 24 hours after learning. Without active repetition, only about 33 percent remains after 24 hours, about 25 percent after a week, and typically 21 to 30 percent after 30 days. Brandon Hall Group has replicated this curve specifically for corporate training contexts and arrives at comparable figures, with variation depending on format and follow-up activity.

The curve has one important property that gets lost in every training discussion. It cannot be beaten with better content. A more brilliant trainer, a more modern slide deck, a more interactive format flatten the curve marginally, but they do not change the underlying equation. Only one thing changes the curve: repeated application in realistic contexts, with feedback.

Repeated application extends the content's half-life. Repeated application with immediate feedback extends it considerably further. Repeated application with immediate feedback and a defined excellence standard extends it far enough that knowledge turns into skill. These three stages, application, feedback, standard, are the mechanics of behavior change. They are structurally absent from a classic workshop format.

Why doesn't classic sales training work?

Classic sales training does not work because it is the wrong format for the change it is supposed to produce. Workshops produce knowledge events, not behavioral competence. Competence only builds through practice in a realistic context, and that practice layer is exactly what is missing from most training architectures. The workshop is not redundant. It is just positioned wrong. It belongs at the start of a long practice sequence, not at the end of a short training sequence.

Three structural deficits explain the gap.

Workshops are events, not systems. They happen at a point in time, usually once or twice a year per rep, and they have no built-in mechanism for repetition. After the last slide, whatever was learned is left to fend for itself. Without a follow-up system, the forgetting curve is unavoidable.

Workshops do not provide objective feedback at the individual level. A trainer with eight to sixteen participants cannot deliver a serious, evidence-based assessment of each individual person. What they deliver is group feedback and subjective impression, both of which have value, but neither replaces individual gap analysis.

Workshops have no connection to the real working day. Thursday's discovery call with the skeptical CFO happens without the trainer watching, without a recording, without evaluation against a standard. Whatever got discussed in the workshop floats in a parallel universe next to the real conversation. A bridge between the two is missing.

Put these three deficits together and it becomes clear why L&D teams have been buying similar workshops in slightly new packaging for years. The format does not change the outcome, because the format cannot change the outcome. Leadership sees activity, the trainer sees a booking, the reps see a break from the phone. What nobody sees is the behavior movement in the customer conversation the following quarter.

What architectural layer is missing between the workshop and the customer conversation?

Between the workshop and the customer conversation, the practice layer is missing: a systematic, repeatable practice stage equipped with individual feedback. In modern training architecture it is called the coaching loop or practice loop, and it is the only layer that can actually bend the forgetting curve. Until roughly 2024 it was not viable at enterprise scale, because it required a 1:1 relationship with a coach per employee. Today it is. Voice-native AI coaching platforms can run this layer per employee, in their own language, with their own personas, against their own scorecard.

Four components make up a working practice layer.

An excellence model. The leader defines what excellent performance in the target situation concretely means, in observable, evidence-citable criteria. A scorecard for discovery quality, a rubric for handling price objections, a standard for negotiations with key accounts. Without this artifact, everything that happens downstream has no point of comparison.

Realistic practice personas. An AI counterpart that sounds and reacts like the real customer or negotiating partner. At Sleak that means voice-native: the rep talks through the entire conversation, and the persona pushes back, raises objections, goes off script, and reacts emotionally. A persona can be a choleric buyer, a fact-driven CFO, a charmingly evasive mid-market executive. More on configuring these in AI Negotiation Training for Procurement.

Immediate, evidence-based feedback. Every practice session ends with a scorecard result and specific quotes from the transcript. Not "that was pretty good," but "value argument at minute 4:32 not followed up, discount discussion at minute 7:15 conceded too early, closing question at minute 12:08 never asked." That specificity is the difference between coaching and affirmation.

Endless repeatability. The rep can practice the same scenario two to five times a week, at different difficulty levels, with different personas, against the same scorecard. Repetition in varied context is the only known method for turning knowledge into skill. In an AI-driven architecture it scales without needing a trainer's calendar.

Together, these four components form what actually bends the forgetting curve. Content gets delivered in workshops, applied and consolidated in the practice layer, and converted into procedural skill through repeated practice. The forgetting curve does not flatten because the workshop got better. It flattens because the practice layer finally exists.

How can you scale sales coaching without hiring more trainers?

Sales coaching scales by shifting the mechanical parts of coaching, observation, scoring, feedback generation, practice-scenario creation, onto an AI platform, and leaving the strategic parts, career development, motivation, cultural embedding, with the human sales manager. The model relieves managers rather than replacing them, and enables a depth of coaching per rep that would be unaffordable with classic 1:1 structures. Scaling coaching was a financial problem for decades, because a coach-to-rep ratio of 1:8 or 1:12 was a hard limit. AI coaching architectures remove that limit.

The economics are simple. A classic sales trainer costs 250 to 600 euros an hour in the DACH market. A twelve-week coaching program with weekly 1:1 sessions costs between 3,000 and 7,000 euros per rep. An AI coaching program on a voice-native platform costs, at 2026 DACH market pricing, between 600 and 1,800 euros per rep per year, with unlimited practice frequency. At 100 reps, the model saves between 240,000 and 520,000 euros a year, with greater coaching depth per person.

What matters is the shift in the sales manager's role. The platform gives the manager the data that was previously missing, individual scorecard movement, behavior-specific feedback, trend analytics. What the manager no longer has to do is mechanically observe every call. What they do instead is have a strategic coaching conversation based on the data. The 1:1 turns from a status check into a development conversation.

Managers who understand this shift get better. Managers who read it as an added burden struggle with the rollout. The cultural work on how managers see their own role is the hardest part of a rollout, not the platform configuration. More real-world context on this in the piece on moving from pilot to rollout.

What distinguishes AI coaching from an AI-assisted LMS?

An AI-assisted LMS distributes content with AI assistance and measures consumption. An AI coaching platform measures behavior in real work situations, compares it against an excellence model, and generates targeted practice scenarios for the individual gap. The difference is not cosmetic. It is architectural. Anyone who confuses the two categories buys the wrong product and produces the same forgetting curve, just with more modern slides.

Three architectural decisions distinguish the two categories.

Observation instead of content distribution. An AI coaching platform sees the real conversation, an LMS sees the course completion. Without observing real behavior, no coaching loop is possible, because the connection to the actual working day is missing.

Evaluation against an excellence model. An AI coaching platform compares observation against a scorecard, an LMS has no such point of comparison. Without a scorecard, evaluation is opinion, not coaching.

A closed development cycle. An AI coaching platform turns identified gaps into individual practice scenarios, an LMS points to further modules. The cycle of observation, evaluation, targeted practice, and renewed observation is what actually bends the forgetting curve.

Sleak builds all three components into one connected system. The leader defines an initiative, the platform translates it into a scorecard, the AI coach observes, scores, and generates practice, and the manager sees movement per rep and per team. The model is multi-departmental, so it applies not only to sales but also to procurement, HR, customer success, and compliance.

In practice: what does a modern sales training system actually look like?

A modern sales training system combines classic knowledge transfer with a continuous AI coaching layer. In concrete terms, that means a four-day onboarding workshop for methodology and product understanding at the start, followed by daily 20- to 30-minute AI practice sessions, a weekly manager 1:1 based on scorecard data, and monthly deep-dive sessions on team-wide weak points. The event format is not lost. It gets embedded in a system.

Example: a mid-market B2B SaaS company with 60 AEs in DACH rolls out a new discovery standard in 2025. Instead of a two-day workshop followed by hoping for the best, the company runs the following rollout.

Weeks 1 to 2, excellence definition. Three top AEs and the VP Sales work with the VP Enablement to define the eight criteria of excellent discovery, with example quotes and counter-examples. The result is a scorecard all 60 AEs understand.

Week 3, workshop with methodology rollout. A two-day workshop covers the methodology, the scorecard logic, and the Sleak setup. Reps leave the workshop with a clear understanding of what they will be measured against going forward.

Weeks 4 to 12, practice layer active. Every AE practices a discovery simulation three to five times a week, with different personas (a technically savvy buyer, a process-focused procurement person, a skeptical CFO). Sleak's AI coach scores every session against the scorecard. The VP Sales sees in the dashboard how the team's average movement per scorecard dimension develops.

Week 8, midpoint correction. The data shows the team is weak, across the board, on the "business impact question" dimension. The VP Enablement schedules a 90-minute deep-dive session, followed by two weeks of targeted practice on exactly that dimension. The platform measures the movement.

Week 12, evaluation. The average discovery scorecard score rose from 58 to 71. The win rate on deals where discovery quality landed in the top quartile is 17 percentage points higher than on deals in the bottom quartile. For the first time, the L&D team reports these numbers to leadership. The conversation changes.

Frequently asked questions

Why do sales reps forget what they learn in training?

Because of the Ebbinghaus forgetting curve. Without active repetition in a realistic application context, human memory decays content exponentially, with the steepest loss in the first 24 hours. After 30 days, typically 70 percent is lost, and after 90 days up to 87 percent. The phenomenon is not specific to sales. It is a basic principle of learning.

How can you scale sales coaching?

By shifting the mechanical parts of coaching (observation, scoring, feedback, practice scenarios) onto an AI coaching platform, while sales managers take on the strategic parts (career development, motivation, cultural embedding). That significantly increases coaching depth per rep, without needing manager headcount to grow linearly alongside it.

Does sales training not work at all?

Sales training works when it is embedded in a system with a practice layer. Workshops alone produce knowledge events without lasting behavior change. Only the combination of knowledge transfer, scorecard-based practice with realistic personas, and immediate evidence-based feedback produces measurable behavior change in the customer conversation.

What does a modern AI-driven sales training system cost?

At 2026 DACH market prices, voice-native AI coaching platforms typically run between 600 and 1,800 euros per employee per year, with unlimited practice frequency. That typically replaces 2,000 to 5,000 euros of classic workshop and trainer cost per rep, at significantly greater coaching depth.

Is AI-driven sales training GDPR compliant?

Yes, provided the vendor demonstrates EU hosting, a GDPR Article 28 data processing agreement, pseudonymization of session data against LLM subprocessors, and transparent data processing. Sleak hosts primarily in Frankfurt (Azure), uses AWS and Supabase EU for the application layer, does not train models on customer data, and is EU AI Act compliant. A single subprocessor (ElevenLabs text-to-speech) has transient US touch without storage.

How do you measure employee development with AI?

Through three metrics: capability movement (average scorecard score per person and role over time), behavior trend (frequency and quality of specific target behaviors in real interactions), and revenue link (the correlation between scorecard movement and business outcomes such as win rate or ramp time). These three replace classic LMS metrics such as course completion rate or training NPS.

Which AI coaching platform fits DACH mid-market companies?

A platform that works voice-natively, has a GDPR-compliant EU data architecture, is deployable multi-departmentally (sales, procurement, HR, CS), and allows scorecard transparency with manager customization. Sleak (Munich, voice-native, ISO 27001 certified, EU AI Act compliant) is one example of this category. For a full market overview, see the piece on the best AI sales coaching vendors 2026, the DACH buyer's guide.

Related reading

• The best AI sales coaching vendors 2026: the DACH buyer's guide, a comparison of the leading platforms in the DACH market.
• AI Negotiation Training for Procurement, how the same practice layer works in procurement contexts.
• What is an AI coaching platform for enterprises?, architecture and selection criteria for DACH buyers.
• From pilot to rollout, why 80 percent of AI initiatives at DACH enterprises fail and what a coaching pilot looks like instead.
• Behavioral change at scale: the missing layer in enterprise learning, a related take on the same architectural gap.