July’s Camp.Fire drilled into two of the industry’s hottest, but often conflated, topics: automation versus AI. Guest expert Dennis Colón, who has spent two decades inside adops teams and now builds RPA (Robotic Process Automation) apps at JIFFY.ai, joined Rob to separate the buzz from the business impact.

Automation ≠ AI

Rob framed the puzzle neatly: generative AI might help with creative or decisioning, but an IO (Insertion Order) should either be done correctly or not done at all, no hallucinations allowed. Automation is deterministic code that executes a known workflow, whereas AI is probabilistic, producing answers where judgment or creativity are often required. The distinction matters, because publishers risk choosing the wrong tool (or worse, delaying action) if everything new is marketed as AI.

Preparation: document the work hiding in plain sight

Dennis’s first discovery on most engagements is that the real process often exists only “in someone’s head.” His team now starts every project with a Process Definition Document that captures every step, exception and business rule before a single bot is built:

• First, teams should shadow staff who perform critical but manual tasks. Think trafficking, billing adjustments or programmatic-deal set-up. Then watch the clicks and copy-pastes that never make it into the standard operating procedure.
• Second, quantify what you see. How many tickets flow through the queue, how many systems are touched, and how often do errors trigger make-goods? Those numbers set the baseline for ROI projections.
• Finally, surface the security or compliance checkpoints early. Legal and IT will ask, and having the answers ready removes one of the most common blockers to pilot approval.

Pitch based on money, not magic

Successful automation proposals focus on a single visible bottleneck rather than promising to “automate everything.” Dennis recommended starting with buys that cause the most internal agitation, since a narrow scope sharpens the before-and-after comparison.

Every assumption must translate effort into dollars because finance speaks the language of headcount. Dennis routinely prices pilots at the cost of roughly 1.5 FTEs when three people currently perform the task, and includes benefits overhead, not just salary.

Just as important: position the project as a revenue-enablement play, not a stealth layoff. The more time you can reallocate to optimization, client service and faster billing, the more you satisfy the leadership team’s “do more with the same” mandate while easing job-security fears on the front line.

Navigating sensitivities: politics, compliance, and culture

Even with convincing math, you can expect three common obstacles. Some executives issue blanket mandates to “use AI” without articulating an outcome, whereas reframing the conversation around speed or accuracy lets technology follow need, not the other way round.

Enterprise buyers that traffic competitive data will likely demand SOC-1 or SOC-2 certification plus hard boundaries on where models run and what they can learn, so bring security documentation with you to the first meeting.

And for early pilots, inserting a human-in-the-loop approval step can reassure skeptics that people, not runaway code, remain accountable (this can be relaxed once trust is earned).

Maximize metrics that matter

Our Camp.Fire attendees converged on four core KPIs for automation success:

• Cycle time: From IO to launch or month-end close resonates because faster revenue recognition pleases both clients and finance.
• Error rate: Measured in make-goods or invoice disputes. Hits the margin directly.
• Hours returned to staff: Showcase productivity gains in hours per person per month, and show how these can be redeployed to optimization or upselling.
• Incremental revenue: Demonstrates that automation is additive.

Tracking these numbers before and after each wave of automation arms champions with evidence for the next budget request.

From efficiency to advantage

Freeing teams from copy-paste toil unlocks human skills that machines still lack: relationship nuance, cross-channel strategy, creative problem-solving. As Dennis observed, AI will not grasp that a low-CPM line is part of a strategic “uber-deal” deserving premium placement, but humans will. Automation, therefore, becomes the infrastructure that lets people operate at the top of their job description instead of at the bottom.

Dennis advocates a modular approach: build micro-apps that master a single job, such as populating an OMS (Order Management System) from a PDF IO, or reconciling programmatic discrepancy files. Publishers can then decide whether to scale horizontally (more workflows) or vertically (deeper integration with OMS, CRM, or a data lake).

Automation is no longer optional

Automation projects are not overnight fixes. Complex order-to-cash flows can take months, perhaps even a year, to stabilize. Setting that expectation, while celebrating small wins along the way, keeps the momentum alive. Publishers who start right now will enter 2026 with cleaner data, leaner workflows, and teams focused on value creation instead of ticket triage.

Automation is no longer optional hygiene but a competitive lever. Distinguish it from AI, document ruthlessly, pitch on revenue, and measure relentlessly. Do that, and the next time leadership asks how you plan to “do more with less,” you’ll have an answer grounded in metrics, not magic.

.