Most RevOps teams see executive reporting as a necessary evil. It’s a recurring, labor-intensive exercise that sits somewhere between data entry and a sense of doom because there’s always that one investor with an eagle eye for typos. After the board meeting, the post-mortem is always the same. Too many manual steps, too many hands on the data, and too many opportunities for something to go wrong.
In a recent RevOps Co-op webinar, Camela Thompson, Head of Marketing at RevOps Co-op and longtime revenue operations practitioner, hosted Matt Cox, Director of Strategy & Operations at Cribl, and Nabil Jallouli, CEO and Co-Founder of Rollstack, to explore what it actually takes to automate the last mile of executive reporting — and why getting it right matters far more than most teams realize.
The conversation covered Cribl's AI strategy and how it's applied in practice, the specific pain point of board and business review reporting, the build-versus-buy calculus for operations teams, and the measurable outcomes Cribl has seen since automating their reporting workflows with Rollstack. The numbers at the end of that story are worth sticking around for.
Thompson set the frame for the session before Cox said a word about Cribl, and it's worth holding onto. Coming from a finance background, with early experience consulting for organizations like Microsoft, she's watched how executive perception of operations teams shifts depending on what those teams actually deliver.
"Reports should really be viewed by revenue operators as the minimum viable product. Because I think too many of us focus on whether it's right and what numbers other people want to see, as opposed to what's really critical for revenue operators to be providing from a strategic standpoint — the business insights, which is why it's happening and what we can do to change." — Camela Thompson
This framing is worth internalizing before reading anything else in this piece. The problem with manual reporting isn't just that it takes too long — it's that it consumes the time that should be spent on the analysis layer above the data. When operations teams spend three days pulling numbers into a deck, they have one day left to figure out what the numbers mean. That's backwards, and it's exactly the trap Cox's team at Cribl was caught in.
This connects directly to a pattern RevOps teams encounter constantly: the gap between being a strategic revenue operator and being a ticket-taker. Automating the data-gathering layer isn't just an efficiency play — it's the prerequisite for becoming a genuine business partner.
Cox leads strategy and operations for a company north of $300 million with roughly 1,000 employees, sitting at the intersection of biz ops, cross-functional project management, pricing governance, and company cadence. His team also leads Cribl's internal AI program management — which gave him an unusually clear-eyed perspective on what the AI landscape actually looks like from the inside of a fast-moving company.
His opening position was direct: Cribl isn't going all in on any single AI tool or approach, and that's by design.
"Our perspective at Cribl is that it's too soon to really go all in with one tool or one approach. Like if we made a bet six months ago of, 'Hey, we're gonna go all in on one vendor,' we'd probably be regretting that bet today given how much AI has changed." — Matt Cox
Instead, Cribl has taken a federated approach — functional teams own and drive AI adoption within their domains, a common four-level AI fluency framework exists across the company but is applied differently by each organization, and there's an active culture of building in public through Slack channels, office hours, and lunch-and-learns. The only hard guardrails: customer data must go through security review before being used in any AI system, and a small list of vendors is off-limits.
But the most important governance mechanism isn't a policy — it's a metric. Cribl measures AI value through one lens only: efficiency. Not agent leaderboards, not token consumption, not tool adoption rates.
"What I'm answerable for is a few areas. So the first is I have to meet my same deliverables in twenty percent less time each quarter." — Matt Cox
That twenty percent figure isn't aspirational — it's the standard his leadership holds him to, every quarter. And the mechanism for tracking it is straightforward: time studies, comparing how long tasks took in the prior quarter versus the current one, with agents being built to log their own time savings automatically so the aggregate picture doesn't require constant manual measurement.
This approach to AI strategy — first principles, outcome-focused, skeptical of noise — shapes everything that follows. Cox wasn't looking for the most impressive AI demo. He was looking for a demonstrable difference between life before a tool and life after it.
Jallouli introduced the conceptual framework that underpins Rollstack's entire product thesis, and it's a useful lens for any RevOps team thinking about where time actually goes in the reporting cycle.
The modern data stack is, genuinely, better than it's ever been. BI tools like Tableau, Power BI, Looker, and Omni can produce sophisticated, interactive dashboards that surface exactly the analytics a business needs. The problem isn't the dashboards. The problem is that the people who need the information don't use the dashboards.
According to research Jallouli cited, BI tool adoption across organizations stagnates at around 25% — and that figure may be optimistic. Executives spend an average of ten seconds on a dashboard before churning away. And only 24% of business leaders pass basic data literacy tests — which means the gap between where data lives and where decisions get made isn't just a preference problem, it's a structural one.
"Execs don't go into dashboards. They go into slides, and they go into meetings." — Matt Cox
The result is a fragile manual chain that RevOps teams know intimately: pull from the BI tool, recalculate in a spreadsheet, copy-paste into slides, run a manual QA check, find an error, trace it back, fix it, check again. And repeat, every reporting cycle, for every deck.
This is what Jallouli calls the "last mile problem" — the gap between the data systems where information lives and the delivery formats where decisions actually get made. It's a gap filled entirely by human labor, and it compounds. For teams managing quarterly board decks, monthly business reviews, and customer-facing QBRs simultaneously, the volume of manual work involved isn't just inefficient — it's a structural barrier to doing anything more valuable with that time.
This challenge is well-documented. As the RevOps Co-op blog post on data management for RevOps pros makes clear, the downstream cost of manual data handling isn't just lost time — it's the erosion of trust in the numbers themselves.
Cox walked through the specifics of what the board deck process looked like at Cribl before Rollstack — and the details are recognizable to anyone who has run a similar process at a company with real board accountability.
Cribl's data stack runs Snowflake as the underlying warehouse, DBT for the ELT layer, and Omni as the business intelligence tool (they were on Looker before migrating about a year ago). Charts and dashboards live in Omni. The board deck lives in Google Slides. Getting data from one to the other was entirely manual.
The deck itself — just the data pack, not the discussion slides — runs 80 to 100 slides per quarter. Each slide required pulling charts from Omni, calculating quarter-over-quarter and year-over-year growth metrics separately in Google Sheets, and then manually copying everything into the presentation. North of 50 charts. Roughly 80 metrics and data tiles.
"Because it's so manual, we have as part of our process that our data and analytics team had to go and check every number. Of course, my team, like we don't, we're not gonna send over a poor product to data and analytics. So between step three and four, we're running our own data checks, we're double-checking our numbers, then we go to the data analytics team, they go and do checks, and guess what? They're gonna find errors." — Matt Cox
That multi-layer QA process — necessary precisely because the manual process was error-prone — consumed days. The board timeline is tight. Every day spent compiling data is a day the leadership team doesn't have to review, analyze, and prepare for the actual board conversation. Cox estimated that back when it was just him running the process, the data compilation alone took three days.
And the stakes of getting it wrong aren't abstract. Cribl's board includes investors like Sequoia, whose teams immediately send the deck to analysts who will cross-reference every number across every slide. The example Cox cited: a customer count on slide 13 didn't match a customer count embedded in a chart on slide 67. That kind of inconsistency doesn't just create confusion — it signals to the board that the business isn't being run tightly.
"To our board, this deck is a huge signal of how we're operating as a company. Even if our numbers are great but we've got inconsistencies across the slides, things aren't adding up, they're gonna majorly question how effectively we're running our business." — Matt Cox
This is the real cost of the last mile problem: it's not just time, it's credibility. And credibility, once damaged in a board context, is expensive to rebuild. The RevOps Co-op piece on audit readiness gaps captures a related dynamic — revenue you can report but can't defend creates the same kind of trust erosion, just at a different altitude.
Before committing to Rollstack, Cox's team did what any operationally rigorous team should do: they evaluated whether to build the solution themselves. Cribl has capable data engineers. The technical requirements were clear. And there's always the temptation, especially in an era of LLM-powered coding tools, to just vibe-code something in-house.
Cox walked through exactly why that path wasn't right for this use case.
"Having this functionality built within Cribl wasn't gonna be a unique competitive advantage for us. And so it's like, let's go to the market and find this. We're wasting our time, and this is where we want something off the shelf that we know is gonna work." — Matt Cox
The reasoning follows a clean framework: Does building this ourselves create strategic differentiation? No — every company has this problem; solving it internally gives Cribl no competitive advantage. What's the blast radius if it breaks? High — the board deck isn't the place to be running beta software. What's the switching cost if a custom build fails? Significant, because the executive team will have already adjusted their workflow around it.
A third consideration Cox raised is worth flagging separately: time cost to the data engineering team. When the data engineers are spending time prototyping a slide-sync tool, they're not doing the work that actually advances Cribl's technical roadmap. The opportunity cost of build is often invisible until it isn't.
The decision landed on buy — and within roughly a month of the data engineering team's prototyping effort, the case was clear. Rollstack was connected to the BI tool in minutes, and the board deck workflow was fundamentally changed.
For RevOps teams thinking through similar decisions, the RevOps Co-op blog post on rebuilding a tech stack explores how to approach vendor evaluation without letting switching costs paralyze the process.
The implementation itself was straightforward. Rollstack connects directly to Cribl's BI tool (first Looker, then Omni after their migration) and maps existing charts and data views to specific positions in the Google Slides deck. Once configured, the workflow is: click run, wait ten to fifteen minutes, and 100-plus syncs against live data populate the entire deck automatically — charts, tiles, growth metrics, everything.
"What once was like 80-plus slides, now we get to kind of click and just have updated in a matter of 10, 15 minutes." — Matt Cox
But the more important outcome isn't speed — it's accuracy. And Cox was direct about which metric he's most proud of.
In the three quarters since fully automating the board deck reporting through Rollstack, Cribl has had zero data errors in board presentations.
The path to that outcome required a deliberate land-and-expand approach. Cox didn't give Rollstack access to the entire business upfront. The board deck was the first use case. Once that proof point was established, monthly business reviews followed (a smaller deck, roughly 25 slides). From there, the finance team added their own spreadsheet-based data to the sync process. And recently, a member of the customer experience team reached out about automating QBR preparation.
The expansion happened organically because the pain point is consistent across functions. Everyone who produces data-driven presentations for stakeholders has the same problem. Once one team demonstrates the solution, the conversation becomes easy.
Cox also noted a practical advantage in Rollstack's usage-based pricing model: adding users doesn't require renegotiating a contract or purchasing additional seats. Users can test independently, find value at their own pace, and the cost scales with actual usage. For an internal champion trying to drive adoption across a skeptical organization, that kind of low-friction expansion matters.
Jallouli was careful throughout the session to distinguish between what AI can reliably do in the reporting workflow and what it cannot — a distinction that matters particularly when the output is going to a board, a CFO, or a customer.
Rollstack uses AI, but in a deliberately bounded way. The platform points to governed, curated datasets — pipelines built by data engineers and analysts, organized by topic — rather than crawling a data warehouse looking for patterns. The AI generates narrative language and story frames for the data, handles straightforward mathematical calculations like week-over-week and month-over-month variances, and helps solve what Jallouli called the "cold start problem" — providing an initial draft structure when analysts aren't sure where to begin.
What it doesn't do is make autonomous judgments about why a number changed, or assert causal claims about business dynamics. That's where the human layer remains non-negotiable.
"AI is used, but in a very governed way, ensuring that the output is extremely accurate. And you know, when dealing with things like client presentations, with board deck, with presentations for leadership, you simply cannot afford having the wrong number be on a board presentation. What will you answer if you get asked where does this number come from? You can't say, 'Let me check with Claude.'" — Nabil Jallouli
Cox added a useful illustration from Cribl's own experience with AI-assisted analysis: AI is excellent at identifying which accounts drove a specific change in a metric — "What were the biggest customers that drove this increase in MAUs?" — but struggles with the "why" behind those changes. Ask it to explain causation across a complex data environment and it starts hallucinating. The right division of labor is AI for pattern detection and data retrieval, humans for context, judgment, and narrative.
This framing aligns with the broader RevOps perspective explored in the RevOps Co-op podcast Episode 94: The Boring Work Behind Great AI — the highest-value AI applications are built on top of well-governed data infrastructure, not deployed as a shortcut around it.
While the board deck was the primary Cribl use case, Jallouli broadened the frame to the customer-facing reporting challenge — quarterly business reviews — which represents the other major operational bottleneck for teams managing large customer books.
Rollstack's data on this is striking: customer success managers spend more than 70% of their time on administrative tasks, leaving only 30% for client-facing work. Meanwhile, more than 95% of customer success and account management teams now carry a revenue quota — retention, net dollar retention, expansion. The administrative load is increasing at precisely the moment the revenue accountability is increasing.
QBRs compound the problem because they scale with the size of the book. Ten customers means ten QBRs per reporting cycle. A hundred customers means a hundred. Without automation, the math eventually breaks: either the quality drops, the frequency drops, or the team burns out.
Rollstack's approach to this is the same as the board deck workflow: connect the BI source, create governed templates, use AI to generate narrative structure from the data, and distribute at scale. Jallouli cited Sony Music's The Orchard as an example — a team of a handful of analysts producing personalized performance reports for more than 10,000 artists across platforms like Spotify and YouTube. That's a scale that would be physically impossible with manual processes.
For RevOps teams thinking about the CSM-facing dimension of this problem, the RevOps Co-op piece on managing the sales-to-customer success handoff explores the broader coordination challenge that QBR automation sits within.
Thompson closed the session with a question that gets to the heart of what RevOps professionals should be developing: if AI handles more of the execution layer, what skills become more valuable?
Cox's answer was immediate and unhedged: systems-level thinking.
"Anyone can build an agent, but to build an effective agent, you have to have that first principles thinking and kind of that systems level thinking. And so, yeah, with LLMs, like some of the technical know-how of like writing a script, you can get that done, but I don't think you can replace the level of, of systems or first principles thinking to really solving what you want AI to do for you in a certain use case or process improvement." — Matt Cox
Jallouli pointed to a more tactical near-term skill for RevOps specifically: using LLMs to accelerate Salesforce development. Because Salesforce schemas are structured and governed, and the translation from business requirements to Salesforce implementation follows a logic similar to code, LLMs can compress that development cycle meaningfully — provided the analyst supplies the right schema context and specification inputs.
But both speakers converged on the same fundamental point: AI is not replacing the judgment layer. It's making the execution layer faster. The teams that benefit most are the ones that can clearly define what they want AI to do — not because they handed the problem to a tool, but because they understood the problem well enough to specify a solution.
The throughline of this session is one that Thompson named at the very beginning: the value of operations teams is not the data they produce. It's the insight they extract from it. Automation doesn't threaten that value — it creates the conditions for it to exist. Teams that are still spending three days pulling data for a board deck don't have three days to think about what the data means. That's the real cost of the last mile problem, and it's the first thing worth fixing.
Learn more about how Rollstack helps revenue operations and analytics teams automate the last mile of executive reporting — from board decks and monthly business reviews to customer QBRs — by connecting BI tools directly to the slides, documents, and emails where decisions actually get made.
Check out our blog, join our community and subscribe to our YouTube Channel for more insights.
