At VB Transform 2026, Zillow’s engineering chief made a point that most consulting firms miss when they start thinking about AI: the ROI numbers only hold up if you measure your baseline before you build anything.

That’s not a technical detail. It’s the difference between a deployment that pays for itself in eight weeks and one that becomes a sunk cost you can’t quantify.

Most consulting firms I talk to know their win rate and their average engagement size. They don’t know how many hours a senior associate spends building a pitch deck from scratch, or how much time the team burns re-researching the same industry twice in three months. When you deploy an AI agent without those numbers, you’re guessing at the return.

This article walks through what baseline measurement looks like for consulting firms, which costs to track, and how to set up an AI deployment so the ROI is real from day one.

Why Baseline Measurement Matters More Than the AI Itself

Here’s what happens when you skip the baseline. You deploy a proposal generation agent. It works. Partners like it. Three months later, someone asks if it’s worth the cost, and you say “probably” because you don’t have the before number.

The agent might be saving 15 hours per proposal. It might be saving three. You can’t tell the difference, so you can’t make the case for expanding it or cutting it.

Zillow’s point is that AI ROI is a comparison. If you don’t measure the manual process first, you’re comparing the new system to a feeling, not a fact.

For consulting firms, the manual process has three cost layers that most people underestimate:

1. Direct labor cost. A senior consultant at $180K loaded costs about $90 per hour. If they spend 25 hours building a proposal, that’s $2,250 in direct cost before you count the opportunity cost of billable work they didn’t do.

2. Rework and duplication. Research that gets done twice. Slides that get rebuilt because no one can find the original. Knowledge that lives in someone’s head instead of the firm’s systems. These costs compound across the team.

3. Speed-to-market delay. A proposal that takes three weeks instead of one means you’re pitching later than the competitor who moved faster. That’s not a line item, but it’s a real cost.

Most firms track the first one loosely. Almost none track the second or third, which means they’re measuring 30% of the problem.

The Three Processes Consulting Firms Should Baseline First

You don’t need to measure everything. You need to measure the processes where AI agents deliver the fastest return, and where the manual cost is high enough to matter.

For most consulting firms, that’s three workflows.

Proposal and Pitch Development

This is the easiest one to baseline because it’s episodic. Every proposal is a discrete event with a start and end time.

Track these numbers for the next five proposals your firm produces:

• Hours spent by role (partner, senior consultant, analyst).
• Time from RFP receipt to submission.
• Number of past proposals referenced.
• Number of custom slides built from scratch.
• Win rate and average deal size (you probably have this already).

The typical consulting firm spends 20 to 40 hours on a major proposal. If your win rate is 30%, you’re spending 70 to 130 hours of labor to close one deal. At blended rates, that’s $7,000 to $12,000 in cost-of-sale before travel, before meetings, before anything billable starts.

A Proposal Generation Agent cuts that time by 60% to 75% because it pulls past proposals, case studies, and pricing into a tailored first draft. The partner still reviews and adjusts, but the build time drops from 25 hours to eight.

If you don’t measure the 25 hours before you deploy the agent, you can’t prove the eight-hour number means anything.

Research and Synthesis at Engagement Start

Most consulting engagements start the same way. Someone on the team spends two weeks reading industry reports, pulling competitor financials, summarizing market trends, and building a situation analysis.

That work is valuable, but it’s also repeated. If your firm does three healthcare strategy projects in a year, you’re doing healthcare market research three times, and most of the sources overlap.

Baseline this process by tracking:

• Hours spent on secondary research per engagement.
• Number of sources reviewed.
• Time to produce the first client-ready brief.
• Percentage of research that overlaps with past projects.

A Research Agent doesn’t replace the consultant’s judgment, but it does replace the manual search, read, summarize loop. It runs structured research against your firm’s knowledge base and external sources, produces a one-page brief with citations, and hands it to the consultant for review.

The time savings here aren’t as dramatic as proposal generation, but they’re consistent. If you’re starting 15 engagements a year and each one burns 30 hours of research time, that’s 450 hours, or about $40,000 in loaded cost for work that could be automated.

Knowledge Management and Reuse

This one is harder to baseline because the cost is diffuse. It shows up as repeated work, as questions that take 20 minutes to answer because the information is buried in a deck from two years ago, as IP that gets created twice because no one knew it existed the first time.

Track these proxies:

• Time spent searching for past deliverables or insights.
• Number of “do we have something on this?” questions per week.
• Frequency of rebuilding slides or analyses that already exist.

One consulting firm in our network estimated they were losing eight hours per consultant per month to search and rework. Across a 20-person team, that’s 160 hours a month, or $14,000 in wasted labor.

A Knowledge Agent reads every document the firm produces and answers questions across the entire corpus. It doesn’t eliminate the need for human judgment, but it does eliminate the search time and the risk of recreating work that already exists.

If you don’t measure the search time before you deploy the agent, you won’t know if it’s working.

What Good Baseline Documentation Looks Like

You don’t need a consultant to do this. You need a spreadsheet and two weeks of honest tracking.

Here’s the format we recommend for consulting firms:

Process: Proposal Development

• Date range: [start] to [end]
• Number of proposals tracked: 5
• Average hours per proposal by role:
  • Partner: 8 hours
  • Senior consultant: 18 hours
  • Analyst: 12 hours
• Average time from RFP to submission: 18 days
• Win rate: 30%
• Average deal size: $240K

Cost per proposal:

• Partner: 8 hours × $120/hr = $960
• Senior consultant: 18 hours × $90/hr = $1,620
• Analyst: 12 hours × $60/hr = $720
• Total: $3,300 per proposal

Cost per win: $3,300 ÷ 0.30 = $11,000

That’s your baseline. When you deploy a Proposal Generation Agent and the average drops to 12 hours total, you know you’re saving $2,100 per proposal, or $7,000 per win.

The same format works for research and knowledge management. The key is to measure before you build, so the comparison is real.

How to Set Up an AI Agent Deployment for Measurable ROI

Once you have the baseline, the deployment process changes. You’re not building an agent to see if it works. You’re building it to hit a specific cost reduction or time savings target, and you’re measuring against a known number.

Here’s the sequence we use with consulting firms at the AI audit for consulting firms:

1. Pick one high-cost process. Don’t try to automate everything. Pick the workflow where the manual cost is highest and the agent’s capability is most mature. For most consulting firms, that’s proposal generation.

2. Set a target. If the baseline is 25 hours per proposal, set a target of 10 hours. That’s a 60% reduction, which is realistic for a well-built agent.

3. Deploy the agent and track the same metrics. Use the same spreadsheet. Track the same roles, the same time buckets, the same win rate. The only thing that changes is the process.

4. Compare after 10 cycles. Don’t judge the agent after one proposal. Track 10, then compare the average to the baseline. If you hit the target, expand. If you don’t, adjust.

This approach works because the ROI is a fact, not a guess. You know what you were spending. You know what you’re spending now. The difference is the return.

If you want a structured way to think through which process to baseline first, we built a worksheet that walks through the decision tree. It’s called Deploy Your First Business Agent, and it covers baseline measurement, agent selection, and ROI tracking in a format you can use with your team.

The Omni Approach to Baseline Measurement

When consulting firms come to us for an Omni Audit, the first thing we do is map the baseline. We don’t start with the AI. We start with the manual process, the cost, and the time.

The audit is 60 minutes. We walk through three workflows, document the current state, and identify which agent delivers the fastest return. You leave with three outputs:

• A process map showing where time and cost are leaking.
• A prioritized list of agents ranked by ROI.
• A 90-day deployment plan with measurable targets.

No deck. No follow-up meeting. Just the map, the list, and the plan.

For consulting firms, the workflows we baseline most often are proposal generation, research synthesis, and knowledge management. Those three processes account for 60% to 80% of the recoverable cost in a typical firm, and they’re where agents deliver the most predictable return.

The Proposal Generation Agent pulls past proposals, case studies, and pricing into a tailored draft. The Research Agent runs structured research at the start of every engagement. The Knowledge Agent answers questions across the firm’s entire corpus of deliverables and meeting notes.

All three are part of Omni Ops, and all three are designed to be measured against a baseline, so the ROI is real from day one.

If you want to see what the audit looks like for your firm, book a 60-min Omni Audit. We’ll map the baseline, identify the highest-ROI agent, and hand you a deployment plan you can execute without us.

Why Most Consulting Firms Skip This Step and Regret It

The reason most firms skip baseline measurement is that it feels like extra work. You want to deploy the agent, see if it works, and move on.

The problem is that “see if it works” isn’t a success condition. It’s a guess.

Without the baseline, you can’t tell the difference between an agent that saves 15 hours per proposal and one that saves three. You can’t make the case for expanding it. You can’t defend the cost when someone asks if it’s worth it.

Zillow’s engineering chief is right. The ROI numbers only hold up if you measure before you build. For consulting firms, that means tracking the manual cost of proposals, research, and knowledge management before you deploy a single agent.

It’s not complicated. It’s a spreadsheet and two weeks of honest tracking. But it’s the difference between an AI deployment that pays for itself in eight weeks and one that becomes a line item you can’t justify.

If you want to see how other consulting firms are measuring baseline performance and deploying agents with real ROI, start with our insights on AI strategy or explore the full Omni platform to see which agents fit your firm’s workflow.

The firms that measure first deploy faster, hit their targets more often, and scale their AI investments with confidence. The ones that skip the baseline spend six months guessing.

What to Do Next

If you’re ready to measure your baseline and deploy an agent with real ROI, the next step is an Omni Audit. It’s 60 minutes, three outputs, no deck.

Book my Omni Audit and we’ll map the manual cost of your proposal, research, and knowledge workflows, identify the highest-return agent, and hand you a deployment plan with measurable targets.

If you want to explore the topic further before you book, check out our guides on business AI or read more about how consulting firms are using Omni to recover $80K to $300K per year in wasted labor.

The ROI is real if you measure it. Start with the baseline.