AI will not fix a broken process

By Cotney Consulting Group.  

Fix the process. Standardize the work. Define ownership. Then use AI. 

Artificial intelligence can help contractors work faster, organize information and reduce administrative effort. What it cannot do is repair a business process that was never clearly defined in the first place. 

A contractor with inconsistent estimating, weak project handoffs, poor field documentation or unclear accountability may be tempted to believe that new technology will solve those problems. In reality, AI often exposes the weakness of the process more quickly. 

If the information entering the system is incomplete, inconsistent or inaccurate, the output will reflect the same problems. If responsibilities are unclear, AI will not decide who owns the result. If a company does not know what a successful process should look like, the technology cannot create that discipline on its own. 

AI can accelerate a good operation. It can also accelerate confusion. 

Technology does not create discipline 

Contractors have faced this issue before. A company purchases new estimating software, project management software or a customer relationship management platform with the expectation that performance will improve immediately. The software may be capable, but the company has not agreed on how information should be entered, who is responsible for updating it or how the data will be used. 

When employees use different systems, enter information inconsistently or create workarounds, the technology becomes another layer of confusion rather than a source of control. Some people use the system, while others continue to rely on spreadsheets, text messages or memory. Reports become incomplete. Management loses confidence in the information, and the tool is blamed for problems that began with the process. 

AI will not change that pattern. If the company has not defined its workflow, adding AI may create another point where information is entered, summarized or misunderstood. Technology does not create discipline. It reveals whether discipline already exists. 

Start with the process, not the tool 

Before introducing AI into any area of the business, contractors should be able to explain the current process end-to-end. Who begins the task? What information is required? Where is that information stored? Who reviews it? What decision is made from it? What happens next? 

If the company cannot answer those questions clearly, the process is not ready for automation. A process should be clear enough that two qualified employees can follow it and produce substantially the same result. That does not mean the process has to be perfect. It does mean the contractor should understand how the work is supposed to flow before asking technology to make it faster. 

For example, AI may help summarize a project handoff meeting, but it cannot determine which information the meeting should include. The company must first define the required scope of review, production assumptions, material needs, scheduling concerns, customer expectations and known risks. 

Once that structure exists, AI can help organize the discussion and capture the follow-up. Without the structure, the summary may be polished but incomplete. 

Broken estimating processes produce faster mistakes 

Estimating is one of the areas where contractors are most interested in AI. That interest is understandable. Estimators manage large amounts of information, work under deadlines and make decisions that directly affect profitability. 

AI can assist with document review, scope comparison, clarifying questions and proposal preparation. But it cannot overcome an estimating system that does not use reliable labor data, accurate material pricing or consistent assumptions. 

If production rates are estimated, AI may help produce the estimate faster, but the number is still a guess. If exclusions are not reviewed, the final proposal may sound more professional while still leaving the contractor exposed. If the estimator does not account for access, mobilization, equipment, supervision, cleanup or project risk, technology will not automatically correct the omission. 

Speed has no value if the estimate is built on weak assumptions. The process must first define what information is required, how costs are calculated and who reviews the final number. AI can then support that structure. 

A weak handoff remains a weak handoff 

Many projects lose profitability between when they are sold and when they reach the field. The estimate may contain important assumptions, but those assumptions are never clearly communicated to the project manager or field supervisor. Known risks may remain in the estimator’s notes. Customer commitments may not be documented. Production is left to discover the scope one problem at a time. 

If critical information exists only in the estimator’s memory, it is not part of the company’s operating system. AI cannot repair a handoff that never happens. It may summarize the information provided, but it cannot identify every missing conversation or unwritten commitment. If the estimate, proposal, schedule and customer expectations are not brought together in a structured review, the project team will still begin with incomplete information. 

A strong handoff requires a defined agenda, the right participants and clear accountability for unresolved items. Once those pieces are in place, AI may help capture decisions, assign follow-up tasks and produce a useful project summary. The technology can improve the handoff. It cannot replace the meeting. 

Poor field documentation cannot be automated away. 

Field documentation is another area where AI can provide significant assistance. Voice notes can be organized into daily reports. Photos can be connected withdescriptions. Observations can be summarized for project managers and customers. But AI cannot report what the field never captured.  

If workers do not document delays, changed conditions, extra work, damaged materials or customer requests, the technology has nothing reliable to organize. The field should know exactly what must be documented, when it must be submitted and who reviews it. 

A company that wants better AI-supported documentation must first define what information the field is expected to provide. Crews need to know when photos are required, which delays must be recorded, how extra work is reported and when the information must be submitted. 

The reporting process also has to be realistic. If the company asks field employees to complete long, complicated reports after an exhausting day, compliance will remain inconsistent regardless of the technology. 

A better process captures the right information at the right time with the least unnecessary burden. Because AI can make that information more useful, but it cannot create the original observation. 

AI cannot assign accountability 

Contractors often focus on the task without defining ownership. A report may need to be completed, a change order prepared or a customer updated, but no single person is clearly responsible for ensuring it happens. 

AI can produce reminders and action lists. It can summarize who said what in a meeting. It cannot accept responsibility for the result. Every AI-supported process still needs a person who owns the outcome. 

If a project update is generated, someone must verify the information before it reaches the customer. If AI identifies a possible scope gap, someone must decide whether it affects pricing. If a meeting summary includes unresolved items, someone must follow through. 

Automation does not eliminate accountability. It makes accountability more important. 

Bad data creates confident answers 

AI works from the information it receives. If job-cost data is incomplete, the analysis will be incomplete. If labor hours are charged to the wrong project, the production conclusions will be unreliable. If customer records are outdated, the communication may contain incorrect information. 

If labor hours are not entered correctly, an AI-based productivity analysis may confidently point management in the wrong direction. The output may still look polished. That is what makes poor data especially dangerous in an AI-supported system. The presentation can appear more accurate than the information behind it. 

Contractors should not introduce advanced analysis into an area where basic data is not trusted. Before using AI to identify trends, compare performance or recommend action, the company should review how the underlying information is collected. Timekeeping, material tracking, project status, change orders and customer records must be reasonably accurate and consistently maintained. 

AI does not improve the truth of the data. It improves the speed at which the data can be processed. 

Standardization must come before automation 

Contractors often resist standardization because every project is different. Projects differ, but many of the processes used to manage them should remain consistent. 

Every estimate should address labor, materials, equipment, access, supervision and risk. Every handoff should review scope, schedule, customer commitments and known concerns. Every project should have a defined reporting process. Every closeout should confirm completion, documentation, billing and lessons learned. 

Standardization does not eliminate professional judgment. It ensures that judgment is applied to the same critical areas each time. AI performs best when it is working within a defined structure. A standard checklist, report format or workflow gives the technology a clear framework. 

Without that framework, the company may receive different results from similar situations simply because employees provided information in different ways. Consistency in the process creates consistency in the output. 

Do not automate waste 

One of the most important questions contractors should ask is whether the current task should exist in its present form. Companies sometimes automate reports that no one reads, approvals that add no value or duplicate data entry that developed over time. Making a wasteful task faster does not make it useful. 

Before introducing AI, management should review the process and remove unnecessary steps. Does the information need to be entered more than once? Does every approval serve a purpose? Is the report helping someone make a decision? Is the same update being created in several formats for different people? 

The first improvement may not be automation. It may be elimination. Eliminate what adds no value, simplify what remains and automate only after the process makes sense. A shorter, clearer workflow is easier to train, manage and automate responsibly. 

Management must define the expected result 

Employees cannot use AI consistently if management has not defined what constitutes good work. A project update may be considered complete by one manager and inadequate by another. One estimator may expect a detailed scope review while another relies on a summary. A supervisor may document delays differently from every other supervisor. 

AI cannot produce consistent work from inconsistent expectations. Management must establish the required content, level of detail, review standard and approval process. 

This is especially important when AI creates first drafts. Employees need to understand that a draft is not automatically a finished product. They must know what to verify and who approves the final version. The clearer the expected result, the more useful the technology becomes. 

Fix the process before expanding the technology 

Contractors do not need to delay every use of AI until the entire company is perfect. They should, however, avoid placing AI into a process they do not understand or trust. 

A practical approach is to choose one process, map how it currently works and identify where information is lost, duplicated or delayed. The company can then simplify the workflow, define responsibilities and establish the expected result. 

Only after that should management decide where AI can add value. It may help create a report, summarize a meeting, compare documents or prepare a communication. The use should be specific, controlled and reviewed. 

Once the process works consistently, the company can expand the technology into another area. That approach may feel slower than purchasing several tools at once, but it produces stronger results and less disruption. 

AI rewards operational maturity 

Artificial intelligence will become more common in contractor operations. It will appear inside estimating platforms, project management systems, customer communication tools and field applications. 

The contractors who gain the greatest value from AI will not be the ones who adopt the most tools. They will be the companies that understand how their work should flow, trust the information entering their systems and hold people accountable for the outcome. 

AI can reduce administrative effort, improve consistency and help management see patterns more quickly. But it cannot repair unclear expectations, missing information or weak accountability. 

Fix the process. Standardize the work. Define ownership. Then use AI to build a better system faster and more effectively.