Picture a cluster revenue manager on a Sunday evening. She’s covering five upper-upscale properties and she’s about to spend six hours producing the weekly pickup commentary that ownership will read in eleven minutes on Monday morning.
If you’ve sat through a Monday revenue meeting in the last five years, you know exactly what comes next. The analyst is tired before the week even starts. The narrative drifts in voice from property one to property five. And the hours that should be going into group displacement work and pricing strategy are going into report assembly instead.
In most revenue teams I’ve seen, the breakdown is roughly 60% report production and 40% actual analysis. The role was designed for the inverse.
Before we get into how Claude Skills fix this, one thing needs to be said up front. AI does not rescue bad infrastructure. If your team is running on a legacy property management system that exports messy flat files, a revenue management system whose recommendations no one can interrogate, or a channel manager that’s a day behind reality, you’re not going to get the gains in this article. Fix the foundation first. Hoteliers who skip that step end up with sophisticated tools pointed at unreliable data, which is just a faster way to be wrong.
Assuming the foundation is solid, here’s the real problem. The data isn’t the bottleneck. It’s already sitting in your hotel PMS, your RMS, your channel manager, and three Excel exports. The bottleneck is the analyst hours spent turning that data into a narrative for a specific audience in a specific voice against a specific competitive set.
This is exactly the kind of work Claude Skills were built for. In this guide I’ll walk through what a Skill actually is, the five Skills a revenue management team should build first, and the operational gains you can expect within ninety days. The order matters. The most common reason teams stall after three weeks is building the elegant competitive-set Skill before the unglamorous pickup-commentary Skill.
The constraint is hours, not data, once your stack is right
Three things have converged over the last two years to make the analyst-hour problem urgent. None of them are going away.
First, owner reporting cycles have compressed. Ownership groups that used to accept a ten-business-day close now expect seven. The sharper asset managers want their internal reads in five. The compression isn’t coming from the close itself, which is mostly automated. It’s coming from the commentary, which is still written by hand, late at night, by the analyst with the most context and the least time.
Second, the cluster manager model has taken over. A property that used to have a dedicated revenue manager now shares one across three to six hotels. The work scales with the property count. The hours don’t. Something has to give, and what gives is depth of analysis on the properties that aren’t currently on fire.
Third, analyst compensation has risen faster than revenue per available room in most US markets. The opportunity cost of routing those hours into report assembly instead of group displacement work or channel mix optimization has never been higher.
These three forces compound. If you’ve already invested in good technology, you’re paying for clean data and solid recommendations. The waste that’s left is the gap between the data and the narrative. Skills are designed to close that gap.
A Skill is an SOP written for an AI assistant
The easiest way to understand a Skill is to think about something your hotel already has: a standard operating procedure.
An SOP captures how a task should be done. The steps, the standards, the exceptions, the voice. It exists so that the same task gets done the same way on every shift, regardless of who’s running it.
A Claude Skill is the same idea, written for an AI assistant instead of a human. It’s a folder of instructions, in plain English, that the assistant reads whenever it sees a task the folder is built for. Inside the folder: a short description so the assistant knows when to use it, a longer set of instructions that read like training notes for a new analyst, and any reference material the task needs.
This matters because most AI use in hospitality today is one-off prompting. Your analyst opens ChatGPT or Claude, pastes in a week of pickup data, and types “write me a commentary for this.” The output is generic, because the AI has no context. It doesn’t know your segmentation, your competitive set, the voice your ownership expects, or which variances actually matter. So the analyst either accepts a generic output or spends as much time correcting it as writing from scratch.
A Skill captures the context once, in writing, and the AI uses it every time. The methodology, the voice, the thresholds, the edge cases. All of it gets written down once and version-controlled, instead of being retyped into a chat window every Sunday.
Two quick clarifications. A Skill is not a custom chatbot. It extends the standard Claude assistant your team already uses. And a Skill is not the same as connecting Claude to your PMS or revenue management system. That connection is separate plumbing. The Skill is the methodology you apply to whatever data the plumbing provides. The plumbing is the pipe. The Skill is the recipe.
Build the unglamorous Skill first
The five Skills below are listed in deployment order, not impact order. Pickup commentary goes first because it’s the highest-frequency task with the lowest stakes per output. It’s the right place to learn how Skills behave before the stakes rise.
Skill 1: Weekly pickup commentary, the foundation
This Skill takes your daily flash, your on-the-books versus same-time-last-year numbers, your pace data, and your segmentation pulls, and produces the narrative section of the weekly pickup report.
Since this is the first Skill your team will build, it’s worth showing what actually goes inside the folder. The instructions look something like this:
You are drafting the weekly pickup commentary for a 240-key upper-upscale urban property. The audience is the asset management team, who read this Monday morning before the 9 a.m. portfolio call. They want diagnostic language, not marketing language. Don’t write “pickup is healthy.” Write “transient pickup ran +4.2% to last year, driven by retail segment recovery in the 14-day booking window.”
Competitive set: [five named hotels]. Property positioning: premium, historically indexing 5 to 12 points above the set on revenue per available room.
Flagging thresholds: pickup variance beyond plus or minus 3% gets noted; beyond 5% gets its own sentence; beyond 8% gets a paragraph with hypothesized drivers. Pace deceleration beyond 200 basis points versus last year gets called out; beyond 400 basis points gets escalated to the commercial director.
Structure: transient, group, channel mix, 30-day forward look. 250 to 350 words total.
That’s maybe a quarter of the actual Skill. The rest is examples of past commentary the team has considered well-written, examples of pushback ownership has given, and terminology to use and avoid. Here’s a simple rule: anything the cluster manager would explain to a new analyst on day one belongs inside the Skill.
The output of this Skill also feeds Skill 5 at month-end, with no retyping required.
Skill 2: Group displacement analysis, where consistency beats speed
This Skill standardizes how your team evaluates inbound group leads against the transient demand they’d push out. The math has always been deterministic. The variance comes from analysts applying it slightly differently.
Inside: your wash factor assumptions, your transient rate forecast curve by date, your food and beverage and ancillary spend assumptions per group room night, the rate floor at which a recommendation flips to decline, and the override conditions for strategic accounts where the relationship matters more than the math.
The value isn’t speed. A good analyst does this math quickly. The value is that the same group inquiry produces the same recommendation regardless of which analyst on the cluster picks it up.
Skill 3: Competitive set commentary, separating noise from story
This Skill interprets your weekly competitive set performance: the revenue and rate index movements that show whether you’re gaining or losing share against your named competitors.
Inside: your positioning rationale (premium, mid, or value, plus the actual logic behind that position), the historical share index band you’ve held, the explanations your team has used for previous moves (renovation, channel shift, competitor closure), and the conditions under which a move should be flagged as structural rather than transient.
A 200-basis-point swing in one week is noise. A 200-basis-point swing that holds for six weeks is a story. The Skill knows the difference.
Skill 4: Variance commentary, the section ownership reads first
This Skill drafts your variance-to-forecast and variance-to-budget commentary for month-end and quarter-end. It’s the section ownership reads most carefully and the section analysts find most painful to write.
Inside: your typical variance drivers (group wash, OTA channel mix shift, the relationship between rate discipline and net rate after commission), the thresholds that determine which variances get a sentence versus a paragraph, and the framing your ownership prefers.
Ownership doesn’t want to read that revenue per available room was down 4.2%. They want to read why, diagnostically rather than defensively.
Skill 5: Owner report, translating revenue language into asset-management language
This Skill consolidates the outputs of Skills 1, 3, and 4 into the narrative section of the monthly owner package, translated into the language asset managers actually use rather than the language your revenue team uses internally.
Inside: your management agreement’s preferred report structure, your audience’s vocabulary preferences (asset managers care about profit margin and incremental dollars, not pickup), and the items that always make it in regardless of variance (capital plan touchpoints, market events, competitor moves).
Six hours of Sunday writing, compressed to fifty-five minutes
The task is the weekly pickup commentary across a five-property upper-upscale cluster, every Monday morning.
Before. Your cluster revenue manager starts Sunday at 5 p.m. She pulls the competitive set report, the pickup and pace files from each property, the channel mix exports, the on-the-books versus last-year comparisons. She writes five property narratives in sequence, each 200 to 400 words. The work takes 2.5 hours Sunday evening and another 3.5 hours Monday before 9 a.m.
The narratives vary measurably across the cluster, and not because the properties are that different. The variance comes from the analyst’s energy drifting across six hours of writing. Property one gets her at her sharpest. Property five gets her at hour six.
Ownership reads the five narratives in eleven minutes.
After. The analyst pulls the same source files. She runs the pickup commentary Skill against the first property. The Skill produces a 300-word draft in 90 seconds, structured to the established sections. She edits for 6 to 8 minutes, focused mostly on the forward-looking paragraph where her judgment adds value the Skill can’t replicate. She repeats across the five properties.
Total time: 55 minutes, down from six hours.
The delta. Roughly 4.5 hours per week recaptured per cluster manager. Methodology consistency across the five narratives, audited against a 12-criterion rubric, moves from 50% to 85%. Voice consistency moves from 55% to 90%. At a fully-loaded analyst cost of $95,000, the recaptured hours equal 0.11 full-time equivalents per cluster manager. Across eight cluster managers, that’s nearly a full FTE.
But the bigger number is what those hours go toward. Group displacement analyses that used to take 2 to 4 hours get the attention they deserve. Pricing strategy reviews that happened monthly start happening weekly.
Ninety days, one Skill at a time
First 30 days. Build only Skill 1, applied to one property. Not the cluster. One property. The cluster manager owns the build. Success at day 30 is narrow: the analyst runs the Skill on Sunday’s pickup, edits the draft, and would have been comfortable sending the edited output to ownership.
The most common failure here is scope creep. The team gets excited, decides to build Skills 2 and 3 in parallel, and finishes none of them. Resist the urge.
Days 31 to 60. Roll Skill 1 across the cluster. The methodology is stable now; what changes from property to property is the competitive set, the segmentation, and the voice your ownership expects. Codify those as property-specific reference files inside the Skill folder.
Add Skill 2 in the back half of this phase. Success: two analysts running the same group lead independently produce recommendations within 5% of each other on the dollar exposure.
Days 61 to 90. Add Skills 3 and 5. Establish version control. The governance question most teams underestimate: who owns the Skill update when the competitive set changes mid-year, when the property’s positioning shifts after a renovation, when ownership replaces the asset manager and the preferred reporting style changes with them?
The answer can’t be “everyone” or “no one.” A named owner with a quarterly review cadence is the right model.
The Skill drafts. The revenue manager still decides.
The Skill drafts commentary. It does not make pricing decisions. Your RMS recommendation and your analyst’s override authority are unchanged. Any vendor pitching otherwise deserves the skepticism it earns.
Skills are weak in three situations you’ll see often enough to plan for.
First, genuinely novel market disruption. A citywide event reshuffle, a competitor closure, a weather event. The Skill produces confident-sounding narrative from the inputs it has, but the context for interpretation isn’t in its instructions yet. The analyst overrides more in those weeks, not less.
Second, owner-facing commentary that touches contested ground. Renovation impact attribution. Competitive set composition arguments. Fee structure conversations. The Skill drafts; the human signing the page owns what it says.
Third, the methodology trap. A Skill that hard-codes flawed methodology spreads that flaw across the team with a consistency that didn’t exist before. The same thing that makes the Skill valuable makes regular methodology audits more important after deployment, not less. The quarterly review of the Skill itself, not just its outputs, isn’t optional.
What to do Monday, and what to ignore
This week: pick one property and one report. The weekly pickup commentary on the highest-touch property in your cluster. Build the Skill against last week’s data. Run it against this week’s. Iterate twice.
Within a quarter, your cluster manager’s Sunday should no longer be a report-production shift. The hours you got back should be visibly going into work that affects rate, mix, or channel cost. If those hours have quietly absorbed back into the same workload, you haven’t actually changed anything.
Be skeptical of vendors pitching “AI revenue management” as a replacement for the analyst’s judgment. The category of tool that earns its keep is the one that accelerates the analyst, automates the report-production tax, and stays out of the pricing decision itself. The category that overreaches makes a recommendation it can’t defend when the market does something its training data didn’t anticipate.
The first is worth building this quarter. The second is worth ignoring until it earns the room.
