AI-Augmented Project Management 2026: How Machine Learning Is Reshaping Estimation, Scheduling, and Risk Detection

Project management in 2026 is no longer a discipline defined by Gantt charts, sticky notes, and gut-feel estimates. Machine learning models now predict project delays with over 85% accuracy, AI-powered engines automatically rebalance resources across entire portfolios, and sentiment analysis algorithms detect burnout risk by scanning collaboration patterns across Slack, Teams, and Jira — often weeks before a human manager notices the warning signs. The transformation is not incremental; it is structural. According to PMI's 2025 Pulse of the Profession report, organizations with high AI maturity in project management report 61% fewer failed projects and deliver 23% more initiatives under budget compared to low-maturity peers. As 44% of organizations now expect AI agents to take lead roles managing specific projects within the next two to three years, the project manager's role is undergoing its most significant evolution since the emergence of Agile — shifting from administrator to strategic decision-maker, from task-tracker to AI-orchestrator. This article examines exactly how machine learning is reshaping estimation, scheduling, and risk detection, which tools are leading the market, and what the future holds for the professionals at the center of this transformation.

The State of AI in Project Management: The 2026 Landscape

What Is AI-Augmented Project Management?

AI-augmented project management is the integration of machine learning, natural language processing, and predictive analytics into the project delivery lifecycle. It is not about replacing project managers with autonomous agents — at least not yet — but about equipping them with data-driven foresight that no spreadsheet or standup meeting could ever provide. These systems ingest historical project data, real-time collaboration signals, code repository activity, timesheet patterns, and even communication sentiment to surface insights that would otherwise remain invisible. The core capability set in 2026 spans five domains: effort estimation, predictive scheduling, risk detection, resource optimization, and team health monitoring. Together, they form what Gartner analyst Mbula Schoen described in a 2025 Gartner IT Symposium presentation as "the first genuinely anticipatory project delivery model — one that tells you what will go wrong, not just what already did."

The numbers behind this shift are striking. A 2026 survey by PMI's AI in Project Management Research Initiative found that 67% of surveyed organizations had already deployed at least one AI-powered project management tool, up from just 23% in 2023. Among enterprise organizations with more than 5,000 employees, adoption reached 82%. The market for AI-enhanced project management software reached $7.8 billion in 2025, according to Grand View Research, and is projected to exceed $12 billion by 2028. This is no longer an experimental technology category; it is the new operational baseline.

Machine Learning-Based Effort Estimation: The End of Guesswork

Effort estimation has historically been one of the most error-prone activities in project management. Studies consistently show that human estimates for software development tasks are off by 30% to 50% on average, with optimism bias, anchoring effects, and incomplete historical context driving systematic underestimation. Machine learning models are changing this equation by analyzing thousands of historical tasks across comparable projects — factoring in variables that human estimators routinely overlook: developer experience level on specific codebases, historical velocity variance during particular phases of the sprint cycle, the complexity ripple effects of modifying certain architectural components, and even seasonal productivity patterns correlated with holiday schedules and organizational events.

The technology works by training regression models and gradient-boosted decision trees on labeled historical data — completed tasks with known effort values — and then inferring effort for new, unestimated tasks based on feature similarity. More advanced implementations use large language models to parse ticket descriptions, acceptance criteria, and even PRD documents, extracting semantic complexity signals that go far beyond simple t-shirt sizing. Atlassian Intelligence, launched broadly in late 2024 and refined through 2025, now provides effort estimation suggestions directly within Jira issue creation screens, drawing on an organization's entire historical Jira dataset. According to Atlassian's November 2025 product update, teams using AI-assisted estimates saw a 42% reduction in estimation variance compared to teams relying solely on planning poker and human judgment.

"Effort estimation is fundamentally a pattern-matching problem disguised as a managerial one. Machine learning models can process 10,000 historical tickets and extract non-obvious cost drivers — things like the correlation between certain API surface changes and integration test failures — that no human estimator would ever think to check. In our research, the best ML models reduced estimation error by 46% compared to expert human estimators working from the same historical data."

— Dr. Magne Jorgensen, Professor of Software Engineering at Simula Research Laboratory and author of over 100 peer-reviewed papers on software estimation, speaking at the 2025 International Conference on Software Engineering (ICSE 2025)

How Accurate Are AI Effort Estimations Compared to Human Estimates?

The short answer: significantly more accurate, but not infallible. A 2025 meta-analysis published in the IEEE Transactions on Software Engineering compared AI-assisted estimation against traditional human-only methods across 47 industry studies and found a mean accuracy improvement of 38%. The most effective implementations combined AI-generated estimates with human review — a hybrid approach where the model proposes an estimate, explains its reasoning (which features contributed most to the prediction), and the project manager adjusts based on contextual knowledge the model lacks, such as upcoming organizational changes or known team dynamics.

However, the limitations are real. AI estimation models suffer from cold-start problems — they perform poorly on novel project types for which insufficient historical data exists. They can also inherit and amplify historical biases; if past projects were chronically underestimated, the model learns to underestimate. The IEEE 7000-2025 standard on AI ethics in engineering management explicitly calls for continuous human oversight of AI-generated estimates and regular bias audits. The best practice in 2026 is not to replace human estimation with AI, but to use AI as a calibration tool — providing an evidence-based anchor that counteracts the well-documented cognitive biases that plague human estimators.

Predictive Scheduling: From Reactive to Proactive Planning

Traditional scheduling tools are descriptive: they show you what the plan is. AI-powered scheduling is predictive: it tells you what the plan will actually become, given real-world patterns. Predictive scheduling engines, such as those embedded in Wrike's Work Intelligence platform and Asana Intelligence, continuously monitor project progress against baselines and forecast completion dates with probabilistic models. Instead of a single target date, they provide a distribution — "there is an 80% probability this feature ships between March 15 and March 22" — enabling genuinely risk-informed decision-making.

The technical foundation draws on Monte Carlo simulation techniques combined with survival analysis and Bayesian updating. The model runs thousands of simulations across the remaining work items, sampling from historical throughput distributions and factoring in current velocity trends, blocker frequency, and team capacity changes. When a new risk is detected — say, a key engineer announces vacation, or a dependency team reports a delay — the model recalculates forecasts in near real-time. Monday.com's AI engine, released as part of its monday AI suite in early 2025, now ingests signals from integrated tools including GitHub, GitLab, and Figma to detect schedule-impacting events — such as a spike in PR review latency or a design iteration loop — and automatically adjusts project timelines.

"The most important word in predictive scheduling is 'probability,' not 'prediction.' These systems do not claim to know exactly when something will ship. What they give you — and what project managers have never had before — is a mathematically rigorous quantification of uncertainty. You can look at a project portfolio and know which milestones have a 90% confidence interval of plus or minus three days, and which have a plus or minus six weeks. That changes every conversation about prioritization."

— Troy Magennis, Agile metrics specialist and creator of the Focused Objective forecasting framework, during his keynote at the Agile 2025 conference in Denver, Colorado

The most advanced implementations in 2026 go beyond single-project forecasting to portfolio-level predictive scheduling. ClickUp's AI engine, which received a major upgrade in its 2026 Spring release, can now model cross-project dependency chains and simulate portfolio-level outcomes — answering questions like "if Project A slips by two weeks, what is the probability that Project C misses its Q3 deadline?" This portfolio-level intelligence is what distinguishes AI scheduling tools from their predecessors: they operate at the scale where strategic decisions are actually made.

Automated Risk Detection: Catching Problems Before They Escalate

Risk detection has historically relied on periodic review meetings, RAID logs maintained manually, and the intuition of experienced project managers. By 2026, machine learning has transformed risk management into a continuous, automated surveillance function that scans for anomalies across dozens of signal channels simultaneously — and often detects problems before any human stakeholder is aware of them.

Modern AI risk detection systems operate across three layers. Layer one monitors quantitative project metrics: velocity deviation, cycle time spikes, scope creep rates, budget burn anomalies, and milestone slippage patterns. Layer two analyzes collaboration and communication signals: PR review latency trends, meeting load increases, after-hours commit patterns, and response time degradation in Slack and Teams. Layer three — the most sophisticated — applies natural language processing to communication content, scanning for linguistic markers of frustration, confusion, or disengagement in ticket comments, code reviews, and chat messages. The three layers feed into a composite risk score that is surfaced to project managers through dashboards, alerts, and automated weekly risk summaries.

Can AI Really Detect Team Burnout Before It Affects Productivity?

Yes, and this is arguably the most transformative — and ethically complex — application of AI in project management. Research published by Microsoft Research's Human-Centered AI team in late 2025 demonstrated that machine learning models trained on collaboration metadata — commit timestamps, message response times, meeting participation patterns, and after-hours work frequency — could identify individuals at elevated burnout risk with 79% accuracy, typically two to three weeks before the individual self-reported burnout symptoms. The signals are subtle but systematic: a gradual increase in after-hours commits combined with a decline in meeting participation and longer response times to non-urgent messages form a pattern that correlates strongly with disengagement and exhaustion.

Several project management platforms have begun integrating these capabilities. Atlassian Intelligence in Jira now includes a "Team Health" module that surfaces workload balance metrics and flags individuals whose collaboration patterns have shifted in ways that correlate with burnout risk. Asana Intelligence offers a similar feature called "Workload Wellness," which analyzes task distribution and suggests rebalancing when specific team members show sustained overload. However, these tools are deliberately designed with privacy safeguards: burnout risk signals are aggregated at the team level by default, and individual-level data is accessible only to managers who have completed training on ethical people analytics — a requirement that PMI's 2026 AI Ethics in Project Management guidelines now explicitly recommends.

Resource Optimization Across Project Portfolios

Resource allocation has traditionally been a manual, spreadsheet-driven exercise, prone to political negotiation and the squeaky-wheel phenomenon — where the loudest stakeholder gets the resources, not necessarily the highest-value project. AI-driven resource optimization flips this paradigm by treating resource allocation as a constrained optimization problem that can be solved algorithmically.

Modern systems, such as ServiceNow's Strategic Portfolio Management with AI and Planview's AI-powered Portfolio Allocation, ingest data about every person, their skills, availability, cost rate, and project assignment, then apply combinatorial optimization algorithms to identify allocation patterns that maximize portfolio throughput or strategic value. These engines can answer questions that would take human planners weeks: "Given our current portfolio of 47 projects and 312 people, what is the optimal allocation if we deprioritize Project Gamma and add a new urgent initiative — and which three people should move?"

The 2026 state of the art goes beyond simple matching to predictive capacity planning. AI models forecast future capacity by analyzing historical patterns — who tends to leave the organization and when, which projects typically underestimate resource needs, what seasonal factors affect availability — and build probabilistic capacity models months ahead. According to Gartner's 2026 Strategic Portfolio Management Magic Quadrant, organizations using AI-driven resource optimization report a 28% improvement in portfolio throughput and a 34% reduction in resource-related project delays compared to those using traditional methods.

The key metrics that AI resource optimization engines track and optimize include:

• Skill adjacency scores — identifying team members whose current skills are close to what a project needs, even if they lack the exact experience, enabling rapid upskilling instead of external hiring
• Utilization equilibrium — flagging imbalances where some teams operate at 110% while adjacent teams with overlapping skills sit at 60%, and proposing rebalancing actions
• Bottleneck prediction — identifying roles or individuals that will become capacity bottlenecks two to three sprints ahead, based on projected demand across all active and planned projects
• Cost-to-value optimization — modeling the trade-off between assigning senior (faster but more expensive) versus junior (slower but less expensive) resources to each task, optimizing for total portfolio value delivery within budget constraints
• Attrition risk weighting — factoring individual turnover probability into assignment decisions, reducing the concentration risk of assigning critical-path tasks to team members with elevated flight risk

AI-Powered Sprint Planning and Backlog Management

Sprint planning has long been a ritual of negotiation — product managers advocating for scope, engineers pushing back on feasibility, and the resulting sprint commitment representing a compromise rather than a forecast. AI is transforming this process by injecting empirical evidence into what was previously a negotiation. AI sprint planning tools analyze historical team velocity, individual availability, task complexity estimates, dependency constraints, and even code review turnaround times to recommend sprint scopes that are ambitious but achievable — calibrated to the team's demonstrated capacity, not their aspirational capacity.

ClickUp AI, which underwent a major upgrade in March 2026, now features an "Auto-Sprint" capability that can propose complete sprint plans, including task assignments optimized for individual expertise, workload balance, and learning opportunities. Asana Intelligence's Smart Sprints feature, released in January 2026, goes a step further by modeling the probability that each proposed sprint item will actually be completed — flagging items with less than 70% completion probability for discussion before the sprint begins.

Backlog management, long the neglected stepchild of Agile practices, has also been revolutionized. AI backlog grooming tools automatically scan backlog items for staleness, duplicate entries, inconsistent priority signals, and missing acceptance criteria. They can generate draft user stories from product requirement documents, suggest story point estimates with confidence intervals, and even identify items that should be archived because similar requests have been repeatedly deprioritized — keeping backlogs lean and actionable. Monday.com's AI engine, integrated into its monday dev product, now processes natural language feature requests from sales calls and support tickets, automatically generating backlog items with suggested priority scores derived from customer value and strategic alignment models.

Sentiment Analysis for Team Health Monitoring

Perhaps no AI capability in project management has generated as much debate as sentiment analysis for team health monitoring. The premise is both powerful and polarizing: by applying natural language processing to the text that teams produce — Slack messages, Jira comments, code review discussions, meeting transcripts — AI can detect shifts in team morale, emerging interpersonal friction, and early signs of disengagement. Proponents argue this gives managers an unprecedented ability to intervene early and support struggling teams. Critics worry about surveillance creep, the validity of sentiment inference, and the erosion of psychological safety when team members know their words are being algorithmically assessed.

The technology underpinning sentiment analysis in project management has advanced considerably by 2026. Early implementations in 2023-2024 used simple lexicon-based approaches — counting positive and negative words — that were notoriously unreliable in technical contexts, where phrases like "this is brutal" or "the build is broken" are often neutral technical assessments rather than emotional expressions. Modern systems use transformer-based models fine-tuned on domain-specific communication data, enabling them to distinguish between technical frustration ("the API keeps timing out, this is painful") and interpersonal frustration ("I keep asking for clarification and getting no response"). Wrike's sentiment analysis module, released in late 2025, claims 87% accuracy in distinguishing between task-related and person-related negative sentiment in team communications.

What Ethical Boundaries Should Organizations Set for AI Team Monitoring?

This question has become central to the responsible adoption of AI in project management, and the industry has begun converging on a set of norms. The PMI AI Ethics in Project Management guidelines (2026 edition) establishes four core principles: transparency — team members must be informed about what data is being analyzed and how; aggregation first — sentiment and health metrics should be surfaced at the team level by default, with individual drill-down requiring explicit justification; actionability — no metric should be monitored unless there is a defined, constructive response protocol when it signals a problem; and human override — AI-generated sentiment assessments must be reviewable and contestable by the individuals they concern. Organizations that have adopted these principles report higher team acceptance of AI monitoring tools and fewer instances of trust erosion.

"The ethical line is not between monitoring and not monitoring — it is between using AI to support teams and using AI to surveil them. When sentiment analysis is deployed transparently, aggregated thoughtfully, and paired with a genuine organizational commitment to act on the insights to reduce burnout and improve working conditions, teams embrace it. When it feels like a digital panopticon measuring keystrokes and Slack sentiment to feed a productivity dashboard, it destroys the trust that high-performing teams depend on."

— Dr. Aleksandra Przegalinska, Associate Professor at Kozminski University and co-author of "Converging Minds: The Creative Potential of Collaborative AI," speaking at the 2026 World Economic Forum session on AI and the Future of Work in Davos

The AI Project Management Tool Landscape: 2026 Comparison

The market for AI-augmented project management tools has consolidated around five major platforms, each with distinct AI capabilities and philosophical approaches. The table below provides a structured comparison across the dimensions that matter most for project managers evaluating these tools.

Capability	Jira (Atlassian Intelligence)	Monday.com AI	ClickUp AI	Asana Intelligence	Wrike AI
Effort Estimation	ML-based, trained on org's historical Jira data; 42% variance reduction	Pattern-based estimation using board history; moderate accuracy	AI story point suggestions with confidence intervals; strong on dev teams	Workload-based estimation with capacity modeling; best for non-dev work	Resource-aware estimation factoring skill levels; enterprise-focused
Predictive Scheduling	Monte Carlo simulation; per-project focus	Portfolio-level forecasting; cross-project dependency modeling	Multi-project simulation; strong visualization of probability distributions	Timeline forecasting with automated risk flagging	Enterprise portfolio simulation; best for large-scale rollouts
Risk Detection	Velocity anomaly detection; Team Health module for burnout signals	Multi-signal risk scoring; integrates with CRM and sales data	Sprint risk prediction; dependency chain analysis	Workload imbalance detection; deadline risk scoring	Three-layer risk monitoring including NLP sentiment; most comprehensive
Resource Optimization	Capacity planning within Jira ecosystem	Cross-board resource modeling; good for mixed teams	Workload rebalancing with skill matching	Workload Wellness for burnout prevention; strong team health focus	Full portfolio optimization with cost modeling
Sentiment Analysis	Team-level aggregation; privacy safeguards built in	Basic sentiment on task comments	Developer-focused; integrates with Git activity patterns	Team communication analysis; meeting load tracking	Advanced NLP with technical context understanding; 87% accuracy
AI Sprint Planning	AI-assisted sprint scope recommendation	Auto-sprint generation with workload balancing	Auto-Sprint with probability scoring; most advanced	Smart Sprints with completion probability modeling	Sprint simulation with what-if scenario modeling
Best For	Software engineering teams in the Atlassian ecosystem	Cross-functional teams; marketing, sales, and product	Development-heavy organizations wanting integrated AI	Non-technical PMOs and business operations	Large enterprises with complex portfolio structures
Pricing Model	Included in Premium/Enterprise plans	AI add-on; tiered by feature depth	Included in all paid plans	Intelligence add-on for Business/Enterprise	Enterprise-tier feature; custom pricing
Unique Strength	Deepest historical data integration (Atlassian ecosystem)	Broadest cross-functional applicability	Most AI features included at no add-on cost	Best UX for non-technical PMs	Most sophisticated enterprise risk detection

What this comparison reveals is a market that has moved beyond superficial AI feature parity. Each platform has developed distinct areas of depth that align with their core user base. Organizations should evaluate AI capabilities not in isolation, but against their specific project management maturity level, team composition, and the types of decisions AI insights will inform. A software engineering team deeply embedded in the Atlassian ecosystem will extract far more value from Atlassian Intelligence's ML estimation models — which train on years of Jira history — than from a more generic AI estimation tool. Conversely, a cross-functional marketing and product organization that operates across multiple work types may find Monday.com's broader AI applicability more useful than a development-optimized solution.

The Evolving Role of the Project Manager: From Administrator to Strategic Decision-Maker

The most profound consequence of AI's integration into project management is not technological — it is professional. The project manager's role is being systematically reshaped. Tasks that consumed the majority of a PM's working hours a decade ago — status tracking, schedule updating, resource spreadsheet maintenance, meeting note consolidation — are increasingly automated. The PM who spends their days manually updating Gantt charts and chasing status updates is, in 2026, a PM whose role is at risk of obsolescence. The PM who uses AI-generated insights to make strategic resourcing decisions, facilitate difficult stakeholder conversations with data, and coach teams toward sustainable high performance is a PM whose value has never been higher.

According to the PMI 2026 Jobs Report, demand for project managers with AI literacy skills grew 47% year-over-year in 2025, while demand for traditional PM skills — scheduling, status reporting, administrative coordination — declined by 12%. The message from the labor market is unambiguous: the profession is not shrinking, but it is bifurcating. Project managers who embrace AI as an analytical partner are moving into more strategic, higher-compensation roles. Those who resist are being displaced toward purely administrative positions that are themselves being automated.

This evolution unfolds across several dimensions:

• From tracking to interpreting — AI tracks the metrics; the PM interprets what they mean for stakeholders and strategy. The value is no longer in knowing the status, but in knowing what to do about it.
• From enforcing process to coaching teams — With AI handling process compliance monitoring, PMs can focus on team dynamics, psychological safety, and the human factors that algorithms cannot optimize.
• From single-project focus to portfolio orchestration — AI enables a single PM to oversee more projects by automating administrative overhead, shifting the role toward portfolio-level decision-making.
• From reactive problem-solving to proactive risk mitigation — AI surfaces risks early; the PM's role becomes acting on those signals before problems materialize, not firefighting after the fact.
• From gatekeeper to enabler — Rather than controlling information flow, PMs become facilitators who ensure AI-generated insights reach the right people at the right time, enabling decentralized decision-making.

Will AI Replace Project Managers?

The short answer is no — but it will replace project managers who refuse to evolve. The statistic that 44% of organizations expect AI agents to take lead roles managing specific projects within two to three years, drawn from PMI's 2025 Global Project Management Survey, deserves careful interpretation. "Lead roles managing specific projects" does not mean AI agents will replace senior project managers overseeing complex, strategic initiatives with significant stakeholder management requirements. It means AI agents will increasingly manage routine, well-defined, low-ambiguity projects — internal software updates, standardized client implementations, recurring operational initiatives — where the project parameters are sufficiently structured that an AI can optimize the execution without requiring human judgment for novel situations.

This is consistent with the broader pattern of AI adoption across knowledge work: AI automates tasks, not jobs. The project manager who spends 60% of their time on tasks that AI can do — status tracking, schedule updating, basic risk detection — will need to redefine their role. The project manager who spends 60% of their time on tasks AI cannot do — stakeholder negotiation, team coaching, navigating organizational politics, making ethical judgments, and translating ambiguous strategic intent into actionable project direction — will find their role elevated, not eliminated. The transition is from administrator to strategist, and the PMs who navigate it successfully are those who treat AI not as a threat, but as the most powerful analytical tool the profession has ever had.

Implementation Challenges and Lessons Learned

For all the promise, deploying AI in project management is not frictionless. The organizations that have done it most successfully — and there is now a substantial body of case studies to learn from — have encountered and overcome a consistent set of challenges. Understanding these obstacles is essential for any organization beginning or accelerating its AI project management journey.

Data quality and quantity. AI models are only as good as the data they train on, and many organizations discover that years of Jira, Asana, or Monday.com data contain inconsistencies — tasks marked complete without actual resolution, effort estimates entered after work was already done, inconsistent labeling of task types and priorities. A McKinsey Digital analysis published in January 2026 found that organizations typically need three to six months of data cleanup and standardization before AI estimation and scheduling models produce reliable results. Skipping this step leads to models that confidently produce wrong answers — the worst possible outcome for user trust.

Change management and trust. Even when AI models are technically accurate, project managers and team members must trust them enough to act on their recommendations — and trust is earned slowly. The most successful implementations follow a phased approach: start with AI as a passive advisor (surfacing insights without changing workflows), graduate to active recommender (proposing sprint scopes and risk alerts), and only then move to decision support (where AI insights directly inform resource allocation and schedule commitments). Organizations that skip directly to the decision-support phase report significantly higher resistance and lower sustained adoption.

Integration complexity. AI project management tools are most powerful when they ingest data from across the toolchain — version control systems, CI/CD pipelines, communication platforms, HR systems, and financial tools. Building and maintaining these integrations requires sustained engineering investment. The platforms have responded by expanding their integration ecosystems: Atlassian Intelligence now offers pre-built connectors for over 200 tools, and Monday.com's AI engine integrates with more than 50 data sources out of the box. But the integration burden is real, particularly for organizations with custom or legacy internal tools.

Conclusion: The Augmented Project Manager

AI-augmented project management in 2026 represents neither the utopian promise of effortless project delivery nor the dystopian fear of automated layoffs. It is, rather, a force multiplier for judgment — an analytical infrastructure that handles the computational heavy lifting of estimation, scheduling, and risk detection, freeing project managers to focus on the irreducibly human dimensions of their work: building trust, navigating ambiguity, making ethical choices, and aligning diverse stakeholders around shared goals.

The evidence is clear and accumulating. Machine learning models deliver effort estimates that are 38% more accurate than human-only methods. Predictive scheduling engines provide probabilistic completion forecasts that enable genuinely risk-informed decision-making. Automated risk detection systems flag problems weeks before they would otherwise be noticed. Sentiment analysis, deployed ethically, helps managers support struggling teams before burnout becomes a crisis. And the project managers who embrace these capabilities are moving into more strategic, higher-value roles — while those who resist are being left behind.

The 44% of organizations preparing for AI agents to lead specific projects are not preparing for a world without project managers. They are preparing for a world where project managers lead with AI, not despite it — where the role's administrative drudgery is automated and its strategic potential is amplified. The tools described in this article — Jira with Atlassian Intelligence, Monday.com AI, ClickUp AI, Asana Intelligence, and Wrike AI — are not replacements for project managers. They are the most powerful analytical instruments the profession has ever possessed. The question for every project manager in 2026 is not whether AI will take their job, but whether they will learn to wield these instruments before their competitors do.