Skip to main content

The AI Sales Assistant Bet: Your Reps Were Never Going to Fill In the CRM

CRMKonvo #322 - Digital Assistants for Sales Reps
For more than twenty-five years the industry has told the same story: buy a CRM, train your reps, enforce discipline, and the customer insight will flow, followed by success.

It never worked.

As we discussed in our recent CRMKonvo with Nicolas Christiaen, co-founder and CEO of Donna, the most valuable customer knowledge is still created in cars, factory halls and restaurant back offices. It also still dies there. Donna's answer is an AI colleague that calls the field rep, asks how the meeting went, and then updates the CRM. That’s genuine help. And it is a far more interesting model than most vendors offer. Whether it is a cure depends on questions a buyer needs to ask before signing.

And on whether the timing is right. I presented a prototype similar to Donna to the DSAG forum in 2019 and the response was lukewarm, at best. Is the timing right now?

TL;DR

If you want to watch the full CRMKonvo, please go ahead here (optimized for smartphones) or here (optimized for tablets/computers).

Else, be my guest and continue to read.

Or do both …

The Personal Assistant Nobody Budgeted For

The origin story is telling. Nicolas sat in a taxi next to one of his best sellers, who was loudly dictating CRM updates, presentations, emails and quotes. The person on the other end of the line was Nicolas's own personal assistant, whose job description contained none of that. The top performer had simply outsourced the admin he hated to the nearest competent human.

That anecdote is important because it challenges the usual CRM adoption narrative. We keep treating poor data entry as a discipline problem among weak reps. It isn’t. Case in point: a top performing rep was the worst offender; the only reason his data made it into the system was a stolen assistant. Donna's thesis is to give every field rep, the ones driving from customer to customer, the assistant that only top sellers used to have.

The first product was a WhatsApp conversation. Then voice notes. Then proactive phone calls. Nicolas pointed out that a key lesson was variety. People differ: some reps want a call in the car, some a voice note on the walk to the car, some refuse calls altogether. Hence the "hyper-personalized" Donna with different voices, speeds and channels, plus feedback surveys during the first ninety days, because a habit that does not form in that window does not form at all.

This is the right problem to address. Field reps have rejected every screen we have thrown at them; often rightfully so. A screenless interface that follows the rep's routine is a better answer than yet another mobile CRM app. In brief: the rep was never going to adapt to the system, so the system has to adapt to the rep. it shall serve them, not the other way round.

Capture First, Advise Later

The most interesting moment came when Nicolas described Donna's early failure. The original idea was to read the CRM and tell the rep the next best action. It failed within weeks, because unsurprisingly "the data that was used as a basis was plainly wrong". It was incomplete and full of gaps. So, the team turned the product around: "We become basically the center of the capture". Donna sits on the table as a note taker, calls five minutes after the meeting, and flags holes in the next briefing.

The claim is that customers now see ten, fifteen or twenty times more data entering their system of record than before. And it is cleaner data. This makes determining the next best action become meaningful, and only with this much data does coaching actually make sense.

It is the strongest part of Nicolas’ story; it is also the part that deserves the hardest look. Volume is not quality. A tenfold increase in captured notes can be a tenfold increase in transcribed small talk, misheard product names and confident summaries of things nobody agreed on. Our own transcript of this very episode turned "Donna" into "Donner" and "CRMKonvo" into "CRM Convo". Multiply that across a sales force and your clean CRM becomes a well-populated swamp. Ralf put it kindly when he said Donna leaves the crappy legacy data for the CRM vendors to clean up. However, the simple truth is that not only the new data needs governance from day one, or the flywheel spins in the wrong direction. But it’s a start.

The pivot that Nicolas describes is the lesson for buyers. Any vendor that promises AI recommendations on top of your current CRM data is selling a recommendation engine for fiction.

Faster to the Ceiling, or Faster to Average?

There is another hard question to answer: does an AI assistant make average reps better, or just make everyone average faster? Nicolas answered with a ceiling metaphor. Everyone has a limit, and AI shortens the learning curve: juniors become mid-level sooner, mid-level reps become seniors sooner. "So, we'll just push faster to towards that ceiling."

That is an interesting, yet modest claim, more modest than most vendor pitches. But it sidesteps the deskilling risk. If preparation, debriefs and follow-ups are done for you, the muscle memory built by doing them yourself never forms. Nicolas conceded the point openly.  He also argued that buyers are getting sharper, so lazy and less capable sellers will lose competitive bids.

Then there is homogenization. If the seller at one company and the seller at its competitor use the same agent, trained on similar playbooks, they become interchangeable. Nicolas's answer was three layers of differentiation: the product, the quality of the agents, and the human who builds trust. That holds only if the third layer is still being developed. An organization that cuts coaching budgets because "Donna does coaching now" removes the single one layer its competitors cannot buy.

The Coach in Your Pocket and the Manager Who Wants the Data

On the flip side, Donna already runs role plays grounded in the company's playbooks and tells reps, unprompted, how a meeting went. Reps reportedly like this because it comes from a colleague rather than the boss. That is plausible as feedback without hierarchy is easier to swallow.

The obvious next request is for that same data to feed performance appraisals. Nicolas's answer was configuration: coaching data can be switched on or off by level, and in some countries it never reaches the manager. Still, what is legal, what is legitimate and what should be done are three very different topics.

Configuration is not a policy. The moment a rep suspects the friendly AI colleague is also the manager's informant, the debrief turns into a performance and loses its value. In this case, the data quality story above collapses, because people do not confide in a system that may be used against them.

Agents Selling to Agents, and the Last Five Percent

Looking ahead, Nicolas sketched procurement agents talking to sales agents, settling the RFP matching in minutes, and leaving humans a "fill the gap" meeting for the last five percent of nuance. But then: if sixty to seventy percent of the buying process is already done before a seller is even contacted, a field assistant arrives late to the party. Something has to influence the long list earlier. Nicolas agreed that the human touch moves to the very start and the very end, and predicted that the CRM as we know it, the screen and database we train people on, will disappear into a continuously refreshed customer context.

That prediction is directionally right. It is also convenient for a company whose product is the interface replacing the screen. The system of record does not vanish because nobody looks at it. If at all, it becomes more important, because every agent on both sides of the deal reads from it. His rapid-fire pick for the CRM field to kill was the probability to close, "because everyone fills in a number that is purely based on subjective parameters". Agreed. Sort of, as that number is normally preset by the system, based on experience. And it only works if this number’s replacement is more trustworthy and reliable.

Pragmatic Playbook for Enterprise Sales and CX Buyers

AI sales assistants address a real problem, which is more than most categories can claim. If you plan to invest in one, get one on terms you can verify.

Demand quality metrics, not volume metrics. A vendor reporting ten times more CRM entries is reporting activity. Look at outcome metrics instead. Ask for error rates on extracted fields, how corrections flow back, and who audits the captured data. Whether there has been an uptick in conversions and conversion rates. Higher system adoption. Improved customer (and employee) satisfaction. Run a sample review of a hundred records against the original meetings before you scale.

Decide where coaching data goes before you deploy. Write a policy, agreed with HR and works councils where they exist, that defines whether debrief and coaching data can reach managers or appraisals. Make the vendor show you the configuration that enforces it and tell your reps what it is – even better: show it. An assistant the reps do not trust is an assistant they will feed with bogus information.

Keep humans developing the skills the agent takes over. Use the control-group approach Donna's customers often apply: a group with the assistant and one without, again, measured on outcome metrics like conversion, not just meeting counts. Then protect manager-led coaching budgets, because the human layer is your only differentiator once competitors run the same agent. If the business case depends on removing coaching, it is a cost-cutting case dressed as a productivity case.

The CRM was never short of fields. It was short of people willing to fill them. Assistants have the potential to change this.


 

Comments

Last Year's Top 5 Popular Posts

You are only as good as your customer remembers

As you know, I am very interested in how organizations are using business applications, which problems they do address, and how they review their success. In a next instance of these customer interviews, I had the opportunity to talk with Melissa Gordon , Executive Vice President, Enterprise Solutions at Tidal Basin about their journey with Zoho. You can watch the full interview on YouTube. Tidal Basin is a government contractor that provides various services throughout the government space, including disaster response, technology and financial services, and contact centers. Tidal Basin started with Zoho CRM and was searching for a project management tool in 2019. This was prompted by mainly two drivers. First, employees were asking for tools to help them running their projects. Second, with a focus on organizational growth and bigger projects that involved more people, Tidal Basin wanted to reduce its risk exposure and increase the efficiency of project delivery. This way, the compa...

SAP Draws a Perimeter around Agentic AI and What That Means for the Rest of US

The most consequential enterprise AI governance document published this year arrived in late April with surprisingly little fanfare. SAP's updated API Policy, version 4/2026 , is a short document in plain English. The clause that is most interesting is Section 2.2.2. It restricts how autonomous and generative AI systems are permitted to interact with SAP APIs. Read literally, it has the potential to change the architecture of agentic AI projects across every SAP customer landscape. Read carefully, it is also more interesting than the lock-in headlines suggest. The policy targets a specific category of AI behavior, not AI as such. It connects to commercial mechanics that go well beyond API stability. And the literal text, in its current form, will probably not survive the next two policy revisions intact. There is a lot to unpack. I will walk through what the policy actually says, how the SAP-watching community is reading it, what the rest of the major enterprise vendors are doin...

The Illusion of Value: Why Salesforce’s Agentic Work Unit is the New "Bad Query" of the AI Era

The News On February. 25, 2026, Salesforce announced a pricing and metrics update . During the company’s Q4 FY2026 earnings call, CEO Marc Benio ff, together with CMO Patrick Stokes , unveiled the Agentic Work Unit (AWU). Positioned as a metric to quantify the labor performed by autonomous digital systems, Salesforce defines an AWU as one discrete task accomplished by an AI agent. According to Salesforce, this discrete task represents the exact moment " raw intelligence is converted into real work ". It is not a fixed unit but measured as a processed prompt, a completed reasoning chain, or an invoked tool. Salesforce explicitly designed the AWU to move the industry conversation away from the raw consumption of Large Language Model (LLM) tokens. As Benioff noted, tokens only measure "how much an AI talks," whereas the AWU is intended to measure actual business execution. The scale of this rollout is massive. Salesforce reported that its platform has already processe...

LLM Showdown: Comparing ChatGPT, Gemini, and Grok for Automated News Research

The analyst’s day is full of research. Now, this is the age of AI and AI is here to help, isn’t it? As everyone is talking about copilots and AI agents, why not using the tools at hand to do a little research on research. NB., no one really has a good definition of an AI agent, so this might become an additional topic for research. But I digress. Imagine the following project at hand, which is not only interesting for analysts, btw, but also for a variety of roles in the corporate world. Let’s call it vendor (competitor) monitoring. The job is the following: Research reputable sites for news about a number of vendors, relating to a set of keywords. Reputable sites are high quality news sites, high quality tech publications, high quality analyst sites and, of course the news pages of the vendors in question. Limit the time frame of the search matching to the cadence of my information requirement, e.g., “yesterday” for a daily update or “last week” for a weekly update. Provide a summary ...

Voice AI and Agentic Commerce: Your Personal Agent Rents Its Home From the Other Side of the Force

Every wave of customer-facing technology arrives with the same promise: this time the balance of power shifts toward the customer. As we discussed in our recent CRMKonvo with Dan Miller , founder and analyst emeritus of Opus Research , voice AI is making that promise again, and this time both sides of the counter are armed. The more interesting questions are where the customer's weapon is kept, who pays its rent, and who gets to whisper to it while it works. TL;DR If you want to watch the full CRMKonvo, please go ahead here (optimized for smartphones) or here (optimized for tablets/computers). Else, be my guest and continue to read. Or do both … Automated Voice Was Never Built for You Dan has watched this market almost from its beginning, and he is blunt about its origin. Automated voice entered the contact center " almost entirely for cost savings ". Everything after that was repair work: better speech recognition, task-specific tuning for finance and travel, and...