AI and Construction Planning: Where It Adds Value and Where It Does Not

Artificial intelligence is beginning to change construction planning and project controls. New tools can review large programmes, process project records, automate repetitive tasks and identify patterns that would take a planner many hours to find manually.

For contractors, this creates a genuine opportunity. Better use of project data can reduce administration, identify programme risk earlier and allow planning teams to spend more time supporting delivery.

However, AI does not understand a construction project in the same way as an experienced planner, project manager or site team. It does not automatically know whether a sequence is buildable, whether a subcontractor’s forecast is credible or whether a delay event genuinely affected the critical path.

The value of AI therefore depends on how it is used.

Used properly, it can make a competent planning function faster and more effective. Used without sufficient control, it can generate inaccurate conclusions with an appearance of certainty.

What does AI mean in construction planning?

AI is a broad term covering systems capable of analysing information, recognising patterns, generating content and carrying out tasks that would previously have required significant human input.

Within planning and project controls, this may include tools that can:

• interrogate programme data;
• identify logic and quality issues;
• compare programme revisions;
• automate progress updates;
• review project correspondence;
• assist with reporting;
• identify emerging schedule risks;
• model alternative scenarios; and
• help planners make controlled changes across large programmes.

The Association for Project Management identifies data processing, forecasting and improved decision-making as important potential applications of AI in project delivery. AI can process far more information than a project team could reasonably review manually, but the usefulness of its conclusions still depends on the quality of the underlying data and the judgement applied to the output.

The distinction between automation and professional judgement is important.

Changing the duration of 300 correctly coded activities is an automation task. Deciding whether those durations are realistic, whether the activities should be changed and what the resulting critical-path movement means is a planning judgement.

Why planning is particularly suited to automation

Construction programmes are structured data models.

They contain activities, dates, durations, calendars, relationships, constraints, resources, codes, milestones and work breakdown structures. In many respects, a planner is creating a logical system that describes how the project is intended to be delivered.

This has similarities with software programming. Both disciplines require a complex objective to be broken into smaller components, connected through defined logic and tested to confirm that the resulting system behaves as intended.

The source material provided for this article makes this comparison directly. It describes planners as building work breakdown structures, activity dependencies and programme logic in much the same way that programmers build modules, functions and control flows. It also highlights how programming practices such as automation, version control, testing and reusable components could improve programme production and management.

This is why AI has considerable potential within planning. Many planning activities are repetitive, rules-based and data-heavy, even though the decisions that sit behind them require experience.

How AI can support programme development

Producing a detailed construction programme from first principles requires much more than entering activities into Primavera P6 or Asta Powerproject.

The planner must understand:

• the scope of work;
• quantities and production rates;
• construction methodology;
• procurement requirements;
• access and logistical constraints;
• interfaces between packages;
• design development;
• testing and commissioning;
• contractual dates; and
• the resources available.

AI cannot independently replace this process, but it can assist with several parts of it.

For example, an AI-assisted system may be able to extract information from bills of quantities, specifications, design schedules or previous project programmes. It could then help organise the information into a proposed work breakdown structure or identify activities that may need to be included.

It may also compare the developing programme with similar historic projects and highlight areas where durations, sequence or activity density appear unusual.

This can improve efficiency, particularly during tender periods where a contractor may need to understand a large volume of information within a limited timeframe.

The planner must still determine whether the proposed activities and durations are appropriate. Historic data is useful, but it does not automatically account for the particular constraints, productivity assumptions or risk profile of the current project.

Programme quality assurance

Programme quality reviews are among the clearest current uses of AI and automation.

A large programme may contain thousands of activities and relationships. Manually reviewing every open end, constraint, lag, calendar and float value can be time-consuming and inconsistent.

Automated tools can quickly identify:

• activities without predecessors or successors;
• excessive or inappropriate constraints;
• unusually long durations;
• large positive or negative lags;
• negative float;
• invalid or unusual calendars;
• excessive total float;
• missing activity codes;
• progress recorded out of sequence;
• broken logic;
• inconsistent status; and
• unexplained changes between programme revisions.

These checks can identify symptoms of poor programme construction. They cannot, by themselves, determine whether the programme is correct.

An activity without a successor may be a genuine logic omission, or it may be a legitimate final activity within a particular workstream. A hard constraint may be inappropriate, or it may represent an unavoidable statutory or operational restriction.

The software can identify the issue. The planner must understand it.

This distinction is particularly important under NEC, where the programme must represent the contractor’s plans realistically and demonstrate a practicable approach to delivering the works. A programme does not become realistic simply because it passes an automated schedule-health test.

Reviewing programme revisions

One of the most valuable potential applications of AI is the comparison of programme updates.

Understanding what has changed between two programme submissions is often more important than reviewing the latest programme in isolation.

A revision may contain thousands of individual changes, including:

• revised durations;
• changed logic;
• deleted or added activities;
• modified constraints;
• changed calendars;
• altered progress;
• movement of milestones;
• changes to planned Completion; and
• movement of the critical path.

Some changes are legitimate responses to progress or revised methodology. Others may materially alter the representation of delay or responsibility.

AI-assisted comparison tools can help identify these changes quickly and group them into meaningful categories. Instead of manually reviewing thousands of lines, the planner can focus attention on the changes most likely to affect project delivery or entitlement.

The professional judgement remains in determining:

• why the changes were made;
• whether they reflect the actual site position;
• whether they alter the critical path;
• whether they relate to a compensation event;
• whether responsibility has been shifted; and
• whether the revised programme remains contractually compliant.

Progress updating and reporting

Planning teams spend a significant amount of time collecting progress information and converting it into programme updates and reports.

AI can assist by processing information from:

• site diaries;
• daily reports;
• progress photographs;
• timesheets;
• procurement records;
• meeting minutes;
• BIM models;
• subcontractor returns; and
• document management systems.

It may be possible to identify references to completed, delayed or commenced work and compare them against the programme.

AI can also help generate the initial draft of a programme narrative, summarising changes to progress, critical activities, milestones and emerging risks.

This can save time, but it creates an important risk: project records frequently contain inconsistent or ambiguous information.

A site diary may state that work “commenced” when only preliminary access was achieved. A subcontractor may report an activity as 90% complete even though the remaining 10% controls the release of the next trade. A progress photograph may show physical installation but not testing, inspection or acceptance.

For this reason, AI-derived progress should be treated as a proposed assessment, not as verified programme status.

The planner should confirm progress with the site team and ensure that the update reflects the contractual definition of the relevant activity or milestone.

Risk identification and forecasting

AI is well suited to identifying patterns across large quantities of historic and live project data.

Where suitable data is available, it may highlight activities, work packages or interfaces that show characteristics commonly associated with delay.

Examples could include:

• repeated slippage in design release;
• declining production rates;
• late subcontractor mobilisation;
• procurement durations extending beyond the forecast;
• increasing numbers of unresolved technical queries;
• activity float being progressively consumed;
• milestones being achieved later in each update; and
• recurring delay at particular interfaces.

Construction software providers increasingly promote AI-supported schedule analysis, risk prediction and forecasting. The stated aim is to identify potential schedule, cost and resource problems earlier, allowing the project team to act before the issue becomes critical.

The value lies in directing management attention.

AI may identify that a package has an elevated likelihood of delay, but it cannot manage the subcontractor, redesign the sequence, obtain an approval or resolve an access problem. The project team must still convert the warning into action.

Scenario modelling

AI and automation can make programme scenario testing substantially faster.

A planner may need to assess questions such as:

• What happens if equipment delivery is delayed by four weeks?
• What if the contractor increases the labour available to a particular work package?
• Can the commissioning sequence be resequenced?
• What is the effect of changing the access strategy?
• Which activities become critical if a Key Date moves?
• Can part of the terminal float be protected?
• What is the effect of introducing weekend working?
• Which completion forecast is most sensitive to a particular risk?

Traditionally, each scenario may require the planner to create a programme copy, make a series of manual changes, recalculate the programme and interpret the results.

An AI-assisted planning tool may be able to apply consistent changes across hundreds of activities, recalculate the programme and present the differences within seconds.

For example, a planner could request that all relevant external works activities be extended by 20% to model adverse-weather productivity. The system could identify the correct activities through their work breakdown structure or activity codes, change the durations and recalculate the critical path.

The supplied source describes systems in which instructions written in ordinary language are converted into controlled operations on the programme. These can include bulk duration changes, selection of activities through multiple coding criteria and creation of programme relationships.

This is useful, but the resulting scenario is only as credible as the assumptions applied. Extending every activity by a fixed percentage may be mathematically consistent without being operationally realistic.

AI and NEC programme management

AI could become particularly valuable on NEC projects because the programme has a central contractual role.

A contractor must regularly demonstrate:

• the current sequence and timing of its operations;
• planned Completion;
• the Completion Date;
• Key Dates;
• work by the Client and Others;
• required access and information;
• float;
• time risk allowances; and
• the effect of implemented compensation events.

An AI-assisted review could compare a programme submission against the requirements of Clause 31.2 and the project-specific Scope. It may identify missing contractual dates, absent Client obligations, incomplete coding or changes that have not been properly explained.

It could also help maintain a record of programme comments and track whether each comment has been addressed in the revised submission.

However, AI cannot determine contractual entitlement merely by identifying that a date moved.

For example, a compensation event assessment requires an understanding of:

• the event;
• the contractual risk allocation;
• the dividing date;
• the programme current at that date;
• the actual or forecast effect of the event;
• concurrent or competing delays;
• mitigation;
• float; and
• time risk allowances.

These matters require technical and contractual judgement. The programme calculation is only part of the assessment.

Can AI undertake delay analysis?

AI can support delay analysis, but it should not be treated as an autonomous delay expert.

Delay analysis requires the planner or analyst to establish:

• what happened;
• when it happened;
• why it happened;
• which activities were affected;
• whether those activities were critical;
• what other events were occurring;
• what the contract requires; and
• whether the available evidence supports the conclusion.

AI can assist with the underlying work.

It may review large volumes of:

• correspondence;
• meeting minutes;
• progress reports;
• programme files;
• instructions;
• early warnings;
• compensation events;
• RFIs;
• design records; and
• site diaries.

It may then help create chronologies, identify recurring issues, locate supporting documents or compare the documentary record with the programme.

This can significantly reduce the time required to organise evidence.

The danger arises when the system moves from identifying information to reaching causation conclusions.

A tool may identify that an instruction and a critical activity occurred during the same period. That does not prove that the instruction delayed the activity. The analyst must establish the actual mechanism of delay and test it against the programme and contemporaneous evidence.

Similarly, an AI system may describe the longest path through a programme, but that does not necessarily establish the factual critical path followed during the works.

In forensic delay analysis, a convincing narrative is not enough. The conclusion must be supported by the programme mechanics, project records and appropriate methodology.

AI-generated narratives and reports

Generative AI can produce clear written explanations quickly. This makes it useful for drafting:

• programme narratives;
• executive summaries;
• progress reports;
• risk descriptions;
• meeting notes;
• tender methodologies;
• change summaries; and
• initial delay chronologies.

The planner should remain responsible for the content.

AI systems can state inaccurate information confidently. They may combine separate events, misunderstand contractual terminology, invent missing detail or present assumptions as established facts.

A report can therefore read professionally while being technically wrong.

This risk is particularly serious where the document will be used to support:

• a compensation event;
• an Extension of Time submission;
• adjudication;
• dispute resolution;
• senior management decisions; or
• financial forecasting.

Every material statement should be traceable to the programme, records or a clearly stated professional assessment.

AI should assist the planner in communicating the analysis. It should not become the unidentified author of the opinion.

Data quality remains the limiting factor

The effectiveness of AI depends heavily on the quality and structure of the project data.

Construction projects frequently contain:

• inconsistent activity naming;
• incomplete coding;
• different data-date conventions;
• missing records;
• contradictory progress reports;
• unstructured correspondence;
• poor programme logic;
• disconnected commercial and planning systems; and
• inconsistent subcontractor reporting.

AI does not remove these problems.

It may process poor data faster, but faster processing does not make the result reliable.

Where the baseline programme is defective, the progress records are incomplete and the coding structure changes every month, an AI system may struggle to distinguish genuine project movement from data noise.

A contractor seeking to benefit from AI should therefore begin with basic project-controls discipline:

• consistent programme standards;
• reliable coding;
• controlled programme revisions;
• structured progress collection;
• clear data ownership;
• accurate records; and
• documented planning procedures.

AI works best when the underlying controls environment is already sound.

Confidentiality and project data

Construction programmes and project records can contain commercially sensitive and confidential information.

This may include:

• contract values;
• rates;
• delay positions;
• operational constraints;
• design information;
• security-sensitive details;
• personal information;
• legal advice; and
• commercially sensitive correspondence.

Before uploading programme information or documents to an AI system, the contractor should understand:

• where the information is stored;
• whether it is used to train the system;
• who can access it;
• how long it is retained;
• whether it leaves the relevant jurisdiction;
• whether the Client has authorised its use; and
• whether the system complies with the project’s information-security requirements.

This is particularly important on defence, nuclear, energy and critical-infrastructure projects, where the use of public or externally hosted AI systems may be prohibited.

The fact that an AI tool is convenient does not override contractual confidentiality or information-security obligations.

Professional responsibility does not transfer to the software

In March 2026, the RICS professional standard on responsible use of artificial intelligence came into effect for RICS members and regulated firms. It requires appropriate knowledge, governance, risk management, due diligence, client communication and professional oversight when AI materially influences professional services.

Many construction planners are not chartered surveyors and may not be directly subject to that standard.

However, the underlying principle remains relevant to the wider project-controls profession:

The professional remains responsible for the work, even where AI assisted in producing it.

The source material supplied for this article applies the RICS principles directly to planning and project controls. It argues that professionals should understand the tool’s limitations, govern its use, carry out supplier due diligence, communicate material AI involvement to clients and independently review every output.

That is a sensible standard for any contractor or consultant.

A planner should be able to explain:

• what the system did;
• what information it used;
• which assumptions were applied;
• where the output was independently checked;
• what limitations were identified; and
• who takes responsibility for the conclusion.

Will AI replace construction planners?

AI is likely to change the planner’s role, but it is less likely to remove the need for experienced planners.

The administrative element of planning will continue to reduce. Tasks such as extracting information, running standard checks, comparing programme revisions and producing first-draft reports can increasingly be automated.

The planner’s value will shift further towards:

• understanding construction methodology;
• challenging delivery assumptions;
• leading collaborative planning sessions;
• integrating subcontractor programmes;
• identifying commercial and contractual implications;
• advising the project team;
• negotiating programme acceptance;
• assessing cause and effect; and
• communicating complex issues clearly.

These are not purely data-processing tasks.

They require site experience, technical understanding, commercial awareness and the ability to engage with people who may have conflicting objectives.

The APM similarly describes AI as a means of automating routine work while allowing project professionals to focus more heavily on leadership, stakeholder engagement and higher-value decision-making.

The planner who only updates dates and issues reports is more exposed to automation than the planner who understands how the project should be delivered and can advise the team accordingly.

A practical approach for contractors

Contractors do not need to adopt every new AI product immediately.

A controlled approach is more effective.

Begin by identifying repetitive activities that consume planning time without requiring significant judgement. These may include programme comparisons, quality checks, data classification and first-draft reporting.

Test the tool on completed or non-critical data before using it on live contractual submissions.

Compare the AI output with a manual professional assessment. Identify where it succeeds, where it fails and which checks are required.

Document how the tool may be used and what information must not be entered into it.

Most importantly, ensure that a competent planner remains responsible for reviewing and approving the result.

The correct objective is not to remove the planner from the process. It is to allow the planner to spend less time processing information and more time improving the project.

Conclusion

AI will become an increasingly important part of construction planning and project controls.

Its strongest current applications are likely to include:

• programme quality assurance;
• revision comparison;
• data processing;
• progress reporting;
• risk identification;
• scenario modelling;
• document review; and
• automation of repetitive programme changes.

These capabilities can save significant time and help contractors identify risks earlier.

However, AI cannot determine whether a construction plan is genuinely achievable, manage the site team, understand every contractual implication or establish delay causation without professional interpretation.

A well-managed AI system can process the programme. It cannot take responsibility for it. The contractors that gain the greatest value will not be those that replace planners with software. They will be those that combine efficient technology with experienced planning judgement, reliable project data and clear professional accountability