Utilization is a staffing decision, not a monthly report

Most consulting firms don’t have a utilization problem because partners fail to review a spreadsheet. They have one because the spreadsheet reports the issue after the useful staffing decisions have already passed.

A consultant finishes a project two weeks early. A signed statement of work starts 10 days later than expected. A senior manager is allocated at 110 percent across three engagements because no one can see the combined picture. Meanwhile, a capable consultant sits unassigned because their skills are described differently in three separate project plans.

By the time finance updates the utilization report, bench time has already happened.

For a consulting or advisory firm between $1 million and $25 million in revenue, the practical target is usually 75 to 85 percent billable utilization for delivery staff. The right number varies. A strategy firm that relies heavily on senior-led business development may run lower. A project-based implementation firm may need to run toward the higher end. The important point is that the target needs to account for non-billable work that drives future revenue, including proposal development, account management, training, and building reusable intellectual property.

The trouble starts when the firm treats every non-billable hour as the same thing.

Some non-billable time is necessary. Some is investment. Some is preventable waste.

A partner spending 12 hours preparing for an important client workshop can be appropriate. Three senior people each spending 25 hours rebuilding a proposal that closely resembles work the firm has sold before is different. A consultant doing targeted research for a new sector may improve the quality of an engagement. Five consultants separately searching for the same industry facts because prior research is buried in old folders is an operating problem.

This is where AI dashboards and agents can help. Not by treating people like units on a production line. By giving partners a forward view of capacity, work, skills, and risk early enough to make better decisions.

Why conventional utilization reporting falls short

Most firms pull utilization data from a mix of systems. Timesheets might sit in a professional services automation platform. Project plans live in Excel, Monday, Smartsheet, or a PM tool. Sales forecasts are in a CRM. Staff skills are partly in a resourcing file and partly in partners’ heads.

That creates three blind spots.

First, the data is historical. A month-end utilization figure tells you who was billable last month. It doesn’t tell you that two consultants will be free on the 14th, an extension is doubtful, and a new project requires capabilities the available team doesn’t have.

Second, the forecast is too coarse. “We have $400,000 in pipeline” is not a staffing forecast. You need to know deal probability, likely start dates, expected duration, project roles, delivery phases, and the level of confidence behind each assumption.

Third, operational work gets overlooked. Senior people often do the staffing puzzle manually through Slack messages, status calls, and memory. They spend Friday afternoon asking who is available, what each person is doing, and when a project will actually start. That’s costly work, but it usually isn’t coded as an operational cost.

A good utilization system changes the operating rhythm from retrospective reporting to forward management. It should answer questions such as:

• Who is likely to be on the bench in the next 2, 4, and 8 weeks?
• Which projects are overstaffed, understaffed, or relying on an overloaded key person?
• Which pipeline opportunities would create a capacity gap if they close?
• Which people have relevant prior experience, even if their formal job title doesn’t show it?
• Where is proposal, research, or internal work consuming senior delivery capacity?
• What staffing move improves utilization without putting delivery quality at risk?

Those are questions an AI-supported dashboard can surface every week, sometimes every day.

The 75 to 85 percent target needs context

A target is useful only if the firm understands its components.

Suppose you employ 12 client-facing people. Each has around 1,650 workable hours a year after holidays, leave, internal meetings, and realistic capacity allowances. That gives you roughly 19,800 potential delivery hours.

At 75 percent billable utilization, you are targeting about 14,850 billable hours. At 85 percent, it is roughly 16,830. The difference is 1,980 hours.

At blended realized rates common in smaller specialist firms, that gap can represent meaningful revenue and margin. It can also be the difference between hiring early, delaying a hire, or accepting a project your current team cannot deliver well.

But pushing everyone to 90 percent or more is rarely a clean answer. It leaves no room for proposal work, practice development, mentoring, quality review, and client recovery when an engagement goes off track. It can also create a hidden delivery problem. Consultants with no available capacity can’t help another team, contribute to a pitch, or address an urgent client request.

The objective is not maximum utilization. It is predictable utilization at a sustainable level.

For many firms, the bigger issue is volatility. The average might look acceptable while individual teams swing between 45 percent and 105 percent. One group is on the bench and another is overloaded. That signals a resourcing and information problem, not a shortage of hours.

What an AI utilization dashboard should actually do

An AI dashboard should not be another screen that partners have to remember to check. Its job is to combine the data your firm already has, identify the next staffing decisions, and explain why they matter.

At a practical level, it pulls from five inputs:

1. Confirmed work: active projects, planned phases, budgets, staffing plans, dates, and actual hours.
2. Sales pipeline: opportunity stage, probability, expected close date, likely project start, scope, and expected team shape.
3. People data: availability, current allocations, location or working constraints, grade, rates, skills, and past experience.
4. Operational work: proposal commitments, research activity, internal initiatives, leave, and training.
5. Knowledge signals: prior case studies, project deliverables, industry experience, and credentials that indicate fit for upcoming work.

The dashboard then creates a rolling capacity view. A useful starting horizon is 12 weeks, with a longer 6-month planning view for hiring and major sales opportunities.

The difference is in the recommendations. Instead of showing a red utilisation percentage, it can flag:

• A consultant with a likely 16-hour weekly bench gap beginning in three weeks.
• A project manager assigned to 95 percent client work and 20 percent proposal effort.
• A pipeline deal that, if won, creates a shortage of two change-management consultants in the planned start month.
• A lower-risk staffing option based on work someone has already delivered in a related industry.
• An upcoming proposal that can reuse 70 percent of a prior response rather than needing a fresh draft.

The partner still makes the call. The system makes the trade-offs visible.

If you want to see where this model fits your current systems and team structure, See Omni for consulting firms. The point is not to replace your project leaders. It is to give them a reliable operating view before utilization slips.

How AI predicts bench time before it becomes lost revenue

Bench prediction is a structured forecast, not a guess from the resource manager.

An AI agent starts by reading the expected end dates and remaining effort on active projects. It compares planned hours with actual burn. If a 10-week project has consumed 85 percent of its budget in week seven, the agent knows the person may become available sooner than the formal project end date suggests. If an engagement is underspending and the client has indicated a possible extension, it can present that as a confidence-weighted scenario rather than treating it as confirmed work.

It then layers in pipeline data. This part needs discipline. A sales opportunity should not be counted at 100 percent because it is in the CRM. The model needs your firm’s actual stage definitions, historical conversion experience where available, and partner input on the likely start date.

For example, a $150,000 diagnostic project marked as “verbal yes” may have an 80 percent probability of starting in four weeks. A broad proposal request with no buying process may belong in a lower-confidence scenario. The dashboard can show both:

• Committed capacity plan, based only on signed or highly certain work
• Likely capacity plan, including weighted pipeline
• Upside scenario, showing the staffing consequence if more opportunities close

That is far more useful than arguing about one forecast number.

The agent should also monitor the work around the sale. Major proposals commonly consume 20 to 40 hours of senior time, often more for complex bids. If a partner, director, and manager all have large proposal commitments in the next fortnight, that affects their available delivery capacity. Most resourcing plans ignore it until delivery work starts to slip.

A forward dashboard turns that into a decision. Reuse existing proposal content. Bring in a different subject matter lead. Defer internal work. Protect the time because the bid is strategically important. The right answer depends on the opportunity. You can only choose it when you can see the load.

Optimizing staffing without treating skills as keywords

The usual staffing process is familiar. A project is sold. A partner asks around. Someone searches old spreadsheets or messages a practice lead. The team gets assembled based on availability, familiarity, and who is loudest in the conversation.

That process works until it doesn’t.

It misses people with relevant experience because knowledge is trapped in past project documents. It repeatedly assigns the same dependable staff members. It creates avoidable bench time in one practice while another pays contractors. And it makes succession difficult because client relationships are concentrated in a few people.

An AI staffing agent can create an initial shortlist by looking at more than title and available hours. It can assess:

• Similar industries and client contexts
• Past project scopes and deliverables
• Functional expertise
• Seniority and commercial rate fit
• Current allocation and planned release dates
• Previous work with the proposed engagement lead
• Location, travel, and client constraints
• Development goals where the firm chooses to include them

This doesn’t mean an algorithm decides who serves the client. Consulting work relies on judgment, chemistry, and trust. It means the partner begins with a credible evidence-backed shortlist rather than a search through memory.

The Knowledge Agent (Omni ops) is central here. It can read the decks, documents, and meeting transcripts your firm produces, then answer questions across that body of work. Ask, “Who has delivered operating model work for mid-market manufacturers?” or “Show our strongest examples of post-merger integration in financial services.” The answer can include the relevant people, projects, deliverables, and source material.

This is not just a knowledge-management upgrade. It improves utilization because staffing decisions become less dependent on the few people who remember every past engagement.

You can see how these operational agents connect across Omni ops, particularly where project delivery, sales activity, and internal knowledge need to work from the same source of truth.

Rebalancing workloads needs a weekly operating rhythm

Automation is useful only when it reaches a person who can act.

A practical weekly rhythm might look like this:

Monday: capacity and risk review

The dashboard sends a concise summary to partners and operations. It identifies people expected to fall below the utilization floor in the next 4 weeks, staff who are overallocated, projects with hours trending above or below plan, and pipeline opportunities that require provisional staffing decisions.

The goal is not a lengthy meeting. For a 20-person firm, 30 minutes is often enough if the data has been prepared.

Tuesday: staffing recommendations

The agent proposes specific adjustments. It may recommend moving a consultant from a low-priority internal task to a project that is short of support. It can identify a senior consultant who can review a proposal for two hours rather than asking a partner to rewrite the whole document.

It should show the consequences. A move that solves this week’s bench gap may create a shortage next month. Partners need the rationale, not a black-box instruction.

Wednesday to Friday: exceptions and learning

Project leaders confirm or reject recommendations. Their reasons improve the model. Perhaps a consultant is available on paper but needs protected time for a professional qualification. Perhaps the client needs continuity and cannot accept a staffing change. These are valid constraints.

Over time, the system learns the firm’s practical rules. It should never hide them.

The workflow works best when agents remove the preparation burden around staffing, not when they force leaders into rigid allocations.

Reduce non-billable drag before hiring more people

Managing utilization is not only about finding billable work for people on the bench. It is also about reducing repetitive non-billable effort that consumes delivery capacity.

Take proposals. A firm may have a reasonable win rate but still lose margin because senior people start each bid from a blank document. They hunt for old case studies, reconstruct pricing approaches, ask colleagues for credentials, and rewrite familiar sections.

The Proposal Generation Agent (Omni ops) pulls relevant past proposals, case studies, credentials, pricing components, and delivery approaches into a tailored first draft. The responsible partner still shapes the commercial argument and checks every claim. But the team isn’t spending 20 to 40 hours rebuilding material that already exists.

Research creates the same issue at the start of an engagement. New work often begins with a rush to understand a client’s industry, competitors, financial position, strategic priorities, and current market conditions. Useful work gets repeated because earlier insights are scattered across project folders.

The Research Agent (Omni ops) can run structured industry and company research, produce sources and summaries, and create a one-page brief for the engagement team. That gives consultants a faster starting point and makes their time available for the work clients actually value.

This matters to utilization because not all recovered hours need to be immediately billed. Some can protect quality, improve pursuit work, or give the firm room to build reusable assets. The key is that leaders can make that allocation intentionally.

For more examples of how firms are approaching operational AI, the Enterprise DNA insights library is a useful place to compare use cases without getting lost in generic AI advice.

Measure the right utilization signals

Don’t judge the system only by a firm-wide billable percentage. Track a small set of leading and lagging indicators.

Start with these:

• Forecast utilization by person and practice for the next 2, 4, 8, and 12 weeks
• Confirmed versus weighted-pipeline demand
• Bench hours predicted, then actual bench hours
• Overallocation risk, especially above 95 percent planned capacity
• Staffing changes made before a project starts
• Proposal hours by opportunity and seniority level
• Time to identify a suitable project team
• Reuse of existing research, case studies, and intellectual property
• Gross margin by project, where your data supports it

A simple review question helps: did the dashboard identify a risk early enough for someone to do something about it?

If it simply reports that utilization was low, it is bookkeeping. If it shows that a consultant will be free in 17 days, identifies two possible assignments, and highlights the commercial trade-offs, it is an operating tool.

For firms in this revenue range, preventable leakage across bench time, over-servicing, slow proposal production, and duplicated research can commonly fall within the $80K to $300K annual range. The exact number depends on team size, rates, sales cycle, and how much senior time is tied up in work that should be reusable. An audit should calculate your version of the problem rather than applying a generic benchmark.

Book a 60-min Omni Audit if you want to map the operational work behind your utilization numbers. In 60 minutes, we identify the highest-value workflow, assess the data and system inputs, and outline a practical agent opportunity. There is no presentation deck to sit through.

Start with one staffing decision, not a giant transformation

The sensible first step isn’t connecting every system and trying to automate the entire firm. Start with one recurring decision that currently depends on manual chasing.

For many firms, that is the weekly question: who will be underutilized or overallocated in the next 4 weeks, and what can we do now?

Build a first view from active project plans, timesheets, leave, and the most credible sales pipeline. Set clear confidence rules. Have partners validate the first recommendations. Then add skill matching, proposal load, research work, and knowledge signals as the process becomes trusted.

If you need a practical way to scope that first agent, download Deploy Your First Business Agent. It is designed as a working checklist for identifying the workflow, inputs, owner, decisions, and guardrails before you build.

You can also access the direct worksheet here: Deploy Your First Business Agent download.

The firms that improve utilization consistently don’t just pressure people to fill timesheets. They make better capacity decisions earlier. They stop rebuilding knowledge they already own. They give partners a clear view of where the next problem will appear.

To assess that opportunity in your firm, start with the AI audit for consulting firms. Then Book my Omni Audit when you are ready to put a dollar value and implementation path around it.