Sales operations used to be a back-office function: keep the CRM tidy, build the reports, run the forecast call. In 2026 it’s a strategic engine for revenue growth, with research from Gartner suggesting that companies investing in data-driven sales operations see roughly 15% higher quota attainment and 20% faster cycles. But the bar has risen faster than most teams’ tooling and data foundations have, and a familiar set of problems has hardened into structural ones.
What follows are the five challenges showing up most consistently — in industry benchmarks and, just as tellingly, in the day-to-day conversations we have with sales and ops leaders.
Where useful, we’ve pulled in patterns straight from real buyer and customer calls to show how these play out in practice.
1. Everyone’s racing to “do AI”, but they’re missing the translation layer
Every sales ops team in 2026 has the same to-do list: build automations, deploy agents, understand the GTM more deeply. Adoption is near-universal, with around 87% of sales orgs now using AI in some form. And yet most of that adoption is shallow experimentation rather than embedded workflow change, and only a minority of organizations report measurable bottom-line impact. Why are tools better than ever, but the results not improving?
The reason is rarely the model. It’s that the AI has nothing coherent to reason over. A CRM stores records; it doesn’t explain what they mean. A deal in the CRM is a row with a stage and a dollar value — not “the third conversation with this account, where the champion’s pricing objection from call one resurfaced and still hasn’t been resolved with the economic buyer.” That gap between raw data and operational meaning is the thing almost no one has built, and it’s the thing that determines whether AI does anything useful. Point an agent at scattered, undefined data and it amplifies the mess.
This missing piece has a name, and the company that has made it famous is Palantir. Their core idea — the ontology, or semantic layer — is best understood as a digital twin of the organization that sits between raw data and the applications consuming it. Instead of tables and columns, you work with real-world concepts: customers, deals, objections, value drivers, and the relationships and business logic connecting them. The distinction industry analysts keep drawing is simple: a database stores information, while an ontology explains what it means and how each entity relates to every other — which is precisely what lets software and AI agents reason over relationships rather than just retrieve records. It’s increasingly described as the operating system for enterprise AI agents, the connective tissue that closes the gap between data and reality.
The most striking proof that this layer is the real prize came in June 2026, when Kirkland & Ellis — the world’s largest law firm by revenue — committed $500m to building its own proprietary AI platform rather than simply using tools from legal-AI vendors. The first product, a “fund formation engine,” was built with Palantir. And the reason they reached for Palantir is the tell: the firm partnered with them specifically for their ontological expertise — to build the connective tissue across the fundraising lifecycle, linking funds, obligations, and transaction history into a single operational system, so knowledge that had lived scattered across thousands of lawyers, documents, and spreadsheets could be structured and applied directly in the flow of work. As one law professor put it, if every firm has access to the same off-the-shelf chatbots, the tools stop being a differentiator.
The lesson for sales ops is the same, just at a different scale. You don’t close the AI value gap by buying another point solution; you close it by building (or adopting) the semantic layer that translates your scattered GTM — calls, emails, CRM fields, notetaker transcripts, knowledge base — into a coherent model an agent can actually act on. This is exactly the gap that shows up in real conversations. One leader admitted they had thousands of calls logged in their CRM but no real idea how many were genuine discovery meetings, “because we haven’t, we don’t really interrogate it.” Another described their interactions as scattered across a notetaker, the CRM, a dialer, and messaging apps, with large chunks never captured anywhere — the textbook precondition for AI that underwhelms. Until that translation layer exists, every automation sits on sand.
This is the problem Doris was built to solve. We’re the semantic layer for your go-to-market — an ontology that translates everything scattered across your calls, emails, CRM, and knowledge base into standardized concepts (deals, people, value drivers, objections, pain points) and, crucially, understands how they relate across an entire deal cycle. It’s the Palantir-style ontology approach, purpose-built for revenue teams rather than a $500m custom build. That layer then fuels whatever sits on top: query it directly in Claude or ChatGPT over MCP, pipe it into your existing tools, or use it inside Doris itself.
2. Reps spend most of their time not selling — and ops owns the leak
The most quoted statistic in sales right now is also the most damning: even with AI adoption near-universal, reps still spend around 70% of their time on non-selling tasks — the rest on deal management, admin, context switching, and internal calls.
Sales ops increasingly owns this problem because the leak is operational. It’s what happens between customer meetings, and it’s almost entirely manual in most orgs.
What the benchmarks describe in aggregate, individual reps describe as a daily experience. A recurring theme in real calls is what we call “commitment debt”: a rep finishes four back-to-back meetings, hits 5pm, and has to mentally rewind to the 10am call to reconstruct what they promised each prospect. The first 48 hours after a conversation are the highest-momentum window for follow-up — and that’s exactly when an overloaded rep is least able to act. The downstream cost is missed commitments, lost momentum, and deals that quietly stall into closed-lost.
For sales ops, the implication is that protecting selling time is now a core mandate, not a nice-to-have. Automating post-call admin is where most of the recoverable hours live — as much as 36 hours a week for a team of five — which is why, once the semantic layer is in place, the highest-leverage first move is letting it draft the follow-up, update the CRM fields, and extract every commitment automatically the moment a call ends.
3. Pipeline visibility breaks down between the calls, not within them
Leaders generally know why high-level deals are lost: the stakeholder, the budget, the competitor. What they struggle with is visibility in real time — where every deal actually sits right now, and what needs to happen next. As one ops leader described their situation, the team had a very good read on lost-deal reasons but the real gap “just comes down to live visibility in each weekly pipeline review.”
This is a structural problem with how deal context accumulates. A single deal might span five or more calls, and the insight that matters often only emerges by connecting them and understanding them as a whole, not as individual calls — for example, an objection raised by a champion on call one that isn’t resolved until call three, by which point that champion may not even be in the room.
CRM fields don’t capture that. They capture a stage and a dollar value, and maybe a summary note that says “went quiet.” Another leader was refreshingly blunt about the consequence: “There’s always a couple of flavor-of-the-week deals that everyone gets attracted to, and others suffer as a result.” Without systematic visibility, attention follows noise, not risk.
This is why questions from the CRO like “Why are our deals stalling at X stage?” have become a RevOps staple. The harder challenge is moving from what happened last quarter to what should we do right now, and for whom — active pipeline management rather than pipeline inspection. This is the payoff of a semantic layer that compounds intelligence across calls rather than treating each one in isolation: the moment context from call one and call three is connected in a single model, “why is this deal stalling?” becomes a question you can actually answer, and ask in plain language.
4. Forecasting is getting harder, and AI hasn’t magically fixed it
Forecasting is the function under the most visible pressure. Only about 7% of sales organizations achieve 90%+ forecast accuracy; the median sits around 70–79%, and roughly 69% of sales ops leaders say forecasting is getting harder, not easier. The operational reality is rougher still: in one large study of more than 270,000 closed-won opportunities, only ~28% closed within 5% of their 90-day forecasted amount, and nearly half were off by more than 50%.
AI helps — AI/ML methods reduce variance and deliver a roughly 15–25% accuracy improvement over manual roll-ups, and a majority of companies using AI forecasting report meaningful gains. But the gains are uneven, and the difference almost always traces back to data quality. As one guide put it, most forecasts are wrong because they’re built on optimism and bad data, then run through a model that amplifies both. AI forecasting needs a relentless feedback loop.
Two structural causes compound this:
First, traditional models lean on seasonality and weighted pipeline stages while ignoring pricing changes, competitor moves, and campaigns — so a competitor launching a free tier in October can reduce close rates the model never saw coming.
Second, and more cultural: many teams don’t even measure forecast-to-actual variance with formal error metrics. Teams talk about the forecast constantly but nobody tracks the errors. If you don’t measure accuracy, you can’t improve it.
And it loops back to challenge one: a forecast is only as honest as the deal data feeding it, which is why the structured signal from conversations — not just rep-reported confidence — is what moves accuracy.
5. Tool sprawl and the gap between AI adoption and AI value
Reps now juggle an average of 7–10 tools. Layered on top is an adoption-versus-value gap: around 87% of sales orgs now use some form of AI, yet much of that “adoption” is shallow experimentation rather than embedded workflow change. McKinsey reports that most organizations report regular AI use somewhere, but nearly two-thirds haven’t begun scaling it.
The lesson emerging from the teams getting real value is that AI lands when it shows up inside the rep’s existing workflow, rather than as yet another separate tab to check. And the fragmentation problem compounds the translation-layer gap from challenge one: when context is spread across a notetaker, the CRM, a dialer, and messaging apps with no unifying model, neither humans nor agents can see the whole picture. Consolidation — and a layer that unifies scattered context into something coherent — is becoming the absolute must for AI to deliver what ops teams want from it.
The point isn’t another tool in the stack; it’s the connective layer that makes the stack you already have legible to AI.
Conclusion
These five challenges aren’t independent.
The missing translation layer is why AI underwhelms; lost selling time starves follow-up; fragmented tools destroy visibility; and forecasts inherit all of it.
The teams pulling ahead in 2026 have stopped treating AI as the thing they’re buying and started treating the semantic layer underneath it — the model that turns scattered GTM data into meaning — as the actual product of sales ops.
For a sales ops leader deciding where to spend the next quarter, the sequencing the evidence supports is clear: invest in the translation layer that makes your GTM legible to AI, automate the admin that’s bleeding selling time, turn that structured data into real-time visibility on where deals actually sit, and only then expect AI — in forecasting or anywhere else — to produce the returns.
Doris is the semantic layer for your go-to-market. We build an ontology across your calls, emails, CRM, and knowledge base so your team — and your AI — can finally reason over the whole deal, not just the latest record. If your sales ops roadmap for 2026 starts with “do more with AI,” start one layer down.
Sources: industry benchmarks and reporting from Gartner, McKinsey, Everstage, Prospeo, Salesforce, Palantir, Business Wire, Bloomberg Law, and XANT Labs (2025–2026), combined with anonymized patterns drawn from real sales and operations conversations.