How AI Workflow Automation Is Cutting Operational Costs by 40% in 2026

Operational costs have always been the pressure point that separates businesses that scale from businesses that stall. In 2026, a growing number of companies across the UK and India are reporting cost reductions of 30 to 40% in specific operational areas, not through headcount cuts alone, but through the systematic deployment of AI-powered workflow automation across functions that were previously considered too complex, too variable, or too human-dependent to automate.

This is not the automation of the previous decade, which largely meant rule-based scripts handling repetitive data entry or simple if-then logic. The AI workflow automation of 2026 understands context, processes unstructured information, makes judgment-based decisions within defined parameters, and integrates across the full stack of business tools that modern organisations depend on. The cost impact is substantial and the adoption curve is steepening rapidly.

This guide examines how AI workflow automation works at a practical level, where the 40% cost reduction figure comes from, which business functions are seeing the largest gains, what the implementation challenges look like, and how businesses of all sizes can approach this opportunity without overextending their technical capability or budget.

Table of Contents

1. What Is AI Workflow Automation in 2026
2. Where the 40% Cost Reduction Figure Comes From
3. Customer Service and Support Automation
4. Marketing Operations and Content Workflows
5. Finance, Admin, and Back-Office Automation
6. Sales Processes and CRM Automation
7. HR, Onboarding, and Recruitment Workflows
8. Data Processing and Reporting Automation
9. Supply Chain and Operations Automation
10. Tools and Platforms Powering AI Automation in 2026
11. How to Implement AI Workflow Automation
12. Risks, Challenges, and What to Watch Out For
13. AI Automation for SMEs: Starting Without an Enterprise Budget
14. The UK and India Context: Adoption Patterns and Opportunity
15. How Prabisha Consulting Supports AI Automation Implementation
16. Final Thoughts: Automation as a Strategic Lever, Not a Cost Exercise

1. What Is AI Workflow Automation in 2026

AI workflow automation refers to the use of artificial intelligence to execute, manage, and optimise multi-step business processes with minimal or no human intervention at each step. It is distinct from traditional automation in one critical way: traditional automation follows fixed rules and breaks when it encounters anything outside its defined parameters. AI automation handles variability, interprets unstructured inputs, and adapts its outputs based on context.

The components of an AI workflow

A modern AI workflow typically combines several layers of technology. Trigger mechanisms detect an event that initiates the workflow, whether that is an incoming email, a form submission, a data threshold being crossed, a calendar event, or a customer action. Large language models or task-specific AI models process the content of that trigger, interpret its meaning, classify its type, extract relevant information, and determine the appropriate response or next action. Integration layers connect the AI reasoning to the business tools where action needs to happen, whether that is a CRM, an email platform, a project management tool, a database, an accounting system, or a customer support platform. Output and monitoring layers log what the automation did, flag exceptions for human review, and provide performance data for ongoing optimisation.

What makes 2026 different

The AI automation capabilities available in 2026 are qualitatively different from what existed even two years ago. Language models that can reliably follow complex multi-step instructions, reason about ambiguous situations, and produce output that meets professional quality standards have transformed the range of tasks that can be delegated to automated systems. Combined with mature integration platforms that connect AI capabilities to virtually every business application in common use, the barrier to deploying meaningful automation has dropped from a significant engineering project to a configuration exercise for many use cases.

2. Where the 40% Cost Reduction Figure Comes From

The 40% figure cited in headlines and vendor case studies deserves scrutiny rather than acceptance at face value. It is real, but it requires context to understand where it applies and where it does not.

Function-specific versus business-wide savings

The 40% reduction is typically measured at the level of a specific business function or process, not across an entire organisation's cost base. A customer support team that automates first-line query handling with an AI agent, resolving 60 to 70% of enquiries without human involvement, may see a 40% or greater reduction in the cost per resolved ticket. A finance team that automates invoice processing, reconciliation, and exception flagging may see comparable reductions in processing time and associated labour cost. Applying the 40% figure to total operational expenditure would require automation across virtually every function simultaneously, which is an aspiration rather than a near-term reality for most businesses.

Where the cost savings actually come from

Labour time is the largest component of operational cost savings from AI automation. Tasks that previously required a human to read, interpret, decide, and act are completed by AI in seconds at near-zero marginal cost. Error correction is a secondary but significant cost saving: AI systems performing well-defined tasks with clear parameters make fewer errors than humans performing repetitive work at volume, which reduces the downstream cost of corrections, rework, and customer complaints. Speed-to-action savings are a third component: processes that previously took hours or days because they sat in human queues are completed immediately, which accelerates revenue recognition, improves customer satisfaction, and reduces the working capital tied up in slow processes.

Realistic expectations by business size

Enterprise organisations with high transaction volumes across standardisable processes see the largest absolute cost savings. For SMEs, the cost saving is often less visible as a percentage reduction and more visible as a capacity gain: the same team can handle significantly more volume without hiring, which is a cost avoidance rather than a direct cost reduction but is equally valuable for growing businesses.

3. Customer Service and Support Automation

Customer service is consistently the function where AI automation delivers the fastest and most measurable return on investment. The economics are straightforward: customer service operations scale linearly with customer volume under traditional models, meaning that every period of growth requires proportional headcount growth. AI automation breaks this relationship.

AI agents for first-line support

AI customer service agents in 2026 are substantially more capable than the scripted chatbots of earlier years. Powered by large language models with access to a business's knowledge base, product documentation, order management system, and customer history, they can resolve a wide range of enquiries including order status queries, returns initiation, product questions, account updates, and billing queries without human involvement. Resolution rates of 60 to 75% for first-line enquiries are consistently achievable for businesses with well-structured knowledge bases, with some implementations reporting higher rates for specific query types.

Intelligent triage and routing

For enquiries that do require human handling, AI triage systems classify the query type, assess urgency and sentiment, identify the relevant department or agent skill set, retrieve relevant customer history and context, and route the enquiry with a summary already prepared. This reduces average handle time for human agents by 20 to 35% by eliminating the information gathering stage of each interaction.

Post-interaction automation

After a customer interaction concludes, AI automation handles the follow-up workflow: generating a case summary, updating the CRM record, triggering any required follow-up actions such as refund processing or escalation tickets, sending a satisfaction survey, and flagging patterns across interactions for service improvement review. Each of these steps previously required human time. Aggregated across hundreds or thousands of daily interactions, the time saving is substantial.

4. Marketing Operations and Content Workflows

Marketing operations is one of the most process-intensive functions in a modern business, involving a continuous cycle of content creation, campaign management, audience segmentation, performance reporting, and optimisation. AI automation has penetrated this function deeply, with measurable impact on both output volume and cost per piece of work delivered.

Content production workflows

AI-assisted content production, where a human strategist defines the brief, target audience, and key messages and AI handles initial drafting, formatting, and variant creation, has reduced the time and cost of producing blog posts, email campaigns, social media content, and ad copy by 40 to 60% for many marketing teams. The key distinction between effective and ineffective implementations is editorial oversight: AI-drafted content that is reviewed, enriched with original insight, and approved by someone with genuine subject knowledge produces strong results. AI-generated content published without meaningful human review produces generic output that underperforms both commercially and in search.

Campaign management automation

AI tools integrated with advertising platforms including Google Ads and Meta automate bid management, audience optimisation, and creative testing at a granularity that no human team can match manually. Performance Max campaigns and Meta's Advantage+ systems use AI to dynamically allocate budget, test creative combinations, and optimise targeting in real time. For businesses running paid campaigns, the question in 2026 is not whether to use AI-driven campaign automation but how to configure and govern it effectively.

Personalisation at scale

Email marketing personalisation has moved well beyond first-name insertion. AI automation systems can segment audiences dynamically based on behaviour, send time-optimised messages to individual recipients, adapt content blocks based on purchase history or browsing behaviour, and trigger multi-step nurture sequences based on engagement signals. Marketing teams that have implemented these capabilities consistently report improvements in email open rates, click rates, and conversion rates alongside reductions in the manual segmentation and scheduling work previously required.

5. Finance, Admin, and Back-Office Automation

Finance and administration functions are characterised by high volumes of structured and semi-structured data processing, rule-based decision making, and repetitive manual tasks that are well-suited to automation. In 2026, AI has extended the reach of automation in these functions beyond what was previously possible with purely rule-based Robotic Process Automation (RPA) tools.

Invoice and accounts payable processing

AI-powered accounts payable automation extracts data from invoices regardless of format, whether PDF, scanned paper document, or electronic invoice, matches invoices against purchase orders and delivery records, identifies discrepancies, routes exceptions for human review, and processes approved invoices for payment. End-to-end processing times that previously took three to five business days are reduced to hours, and error rates from manual data entry are largely eliminated. Cost reductions of 60 to 80% per invoice processed are reported by organisations that have moved from manual to AI-automated AP workflows.

Expense management

AI expense management tools capture receipts via mobile, extract line items, classify expenses against policy categories, flag policy violations, and route claims through approval workflows automatically. The administrative burden on both employees submitting expenses and finance teams processing them is reduced significantly, and policy compliance rates improve because AI applies rules consistently rather than selectively.

Financial reporting and reconciliation

Month-end close processes that previously required days of manual reconciliation work can be accelerated substantially with AI tools that match transactions across systems, identify discrepancies, generate variance analyses, and produce draft management accounts. The finance team's role shifts from data gathering and formatting to review, interpretation, and strategic input, which is a better use of qualified finance professionals and reduces dependence on junior headcount for data processing tasks.

6. Sales Processes and CRM Automation

Sales is a function where the human relationship remains central to outcomes, particularly in B2B contexts. AI workflow automation in sales is therefore most effectively deployed in the surrounding operational work that consumes sales team time without directly driving revenue, rather than in the relationship-building activities where human presence is irreplaceable.

Lead qualification and scoring

AI lead scoring models analyse inbound leads against historical conversion data to predict which leads are most likely to convert and at what deal size. Rather than every sales representative spending time qualifying every inbound enquiry manually, AI scoring routes high-probability leads to senior salespeople immediately, places medium-probability leads into nurture sequences, and filters out low-quality leads before they consume any sales time. The result is a higher proportion of sales time spent on qualified opportunities and a measurable improvement in conversion rates from lead to close.

CRM data hygiene and enrichment

CRM data quality degrades continuously as contacts change roles, companies merge, and information becomes outdated. AI tools that automatically verify and enrich CRM records with current information from public sources, flag outdated records for review, and deduplicate contacts reduce the administrative overhead on sales teams and improve the reliability of pipeline reporting and forecasting.

Sales outreach automation

AI-powered outreach tools personalise prospecting emails and follow-up sequences at scale, drawing on publicly available information about the prospect and their company to craft contextually relevant messages. Sequence automation handles follow-up timing, response detection, and handoff to human representatives when a prospect engages. The sales development function, which is heavily dependent on high-volume outreach, sees significant productivity gains from these tools when they are configured with genuine personalisation rather than superficial variable substitution.

7. HR, Onboarding, and Recruitment Workflows

HR operations involve a substantial volume of process-driven work including recruitment coordination, onboarding administration, policy communication, and routine employee enquiry handling. AI automation has made meaningful inroads into each of these areas without replacing the human judgment that the sensitive dimensions of HR require.

Recruitment workflow automation

AI recruitment tools screen CVs against job specifications, score candidates against defined criteria, schedule interviews by coordinating calendar availability across multiple parties, send communications to candidates at each stage, and compile shortlist summaries for hiring managers. The administrative coordination work that previously consumed significant recruiter time is largely automated, allowing recruiters to focus on assessment, relationship building, and hiring decision support.

Onboarding automation

New employee onboarding involves a predictable sequence of tasks: document collection, system access provisioning, induction scheduling, policy acknowledgement, equipment ordering, and introductory communication. AI workflow automation orchestrates these sequences automatically from the moment an offer is accepted, ensuring that new starters arrive ready to work rather than spending their first days chasing paperwork. The reduction in HR administration time per new hire is significant, particularly for organisations with high hiring volumes.

Employee self-service AI

AI-powered HR assistants handle routine employee enquiries about leave balances, payroll, policy questions, and benefits without requiring HR team involvement. These tools, integrated with HR information systems and policy documentation, provide accurate, consistent answers at any hour without consuming HR advisor time on questions that have straightforward answers.

8. Data Processing and Reporting Automation

Every function in a business generates data, and extracting actionable insight from that data has traditionally required skilled analysts spending significant time on collection, cleaning, transformation, and visualisation. AI automation compresses this cycle substantially.

Automated data pipelines

AI-enhanced data pipelines pull data from multiple source systems, clean and normalise it, identify and handle anomalies, and load it into reporting destinations automatically. Processes that previously required manual intervention when data formats changed or source systems were updated can now be handled by AI that adapts to structural changes without breaking the pipeline.

Natural language reporting

AI reporting tools that translate data into natural language narrative summaries are increasingly common in 2026. Rather than distributing a dashboard that stakeholders must interpret themselves, automated systems generate written commentary on performance trends, highlight significant variances, and flag items requiring attention. This reduces both the time analysts spend writing reports and the time stakeholders spend deciphering them.

Anomaly detection and alerting

AI monitoring tools that continuously scan operational data for anomalies, whether unexpected drops in sales, unusual cost variances, inventory discrepancies, or website performance degradation, and alert the relevant team immediately are replacing the manual review processes that previously caught these issues hours or days after they occurred. Earlier detection translates directly into lower cost of remediation.

9. Supply Chain and Operations Automation

Supply chain management involves enormous complexity: demand forecasting, inventory optimisation, supplier coordination, logistics planning, and exception management across chains that may span multiple countries and dozens of partners. AI automation has delivered measurable efficiency gains across all of these dimensions.

Demand forecasting

AI demand forecasting models that incorporate historical sales data, seasonality patterns, promotional calendars, external economic signals, and even weather data produce more accurate forecasts than traditional statistical models, particularly for products with irregular or seasonal demand profiles. More accurate forecasts reduce both stockout costs and the carrying cost of excess inventory, with combined impact that can account for a significant portion of the overall cost reduction figures reported by supply chain operations that have implemented AI forecasting.

Supplier communication automation

Purchase order generation, delivery confirmation, invoice matching, and routine supplier queries can all be handled through automated workflows that reduce the manual coordination burden on procurement teams. AI tools that monitor supplier performance, flag delivery delays against committed dates, and initiate escalation workflows automatically allow procurement managers to focus attention on strategic supplier relationships and exception handling rather than routine transactional coordination.

10. Tools and Platforms Powering AI Automation in 2026

The technology landscape for AI workflow automation has matured substantially. Businesses in 2026 have access to a range of tools that span from no-code platforms accessible to non-technical teams to developer-grade infrastructure for complex custom implementations.

No-code and low-code automation platforms

Make (formerly Integromat), n8n, and Zapier are the three most widely deployed integration and automation platforms for mid-market businesses. All three have integrated AI capabilities that allow workflows to incorporate LLM-based reasoning steps alongside traditional trigger-action logic. n8n in particular has become the preferred platform for businesses that want self-hosted control over their automation infrastructure without the overhead of building from scratch. These platforms connect to hundreds of business applications via pre-built connectors, making it possible to build sophisticated multi-system workflows without writing code.

AI agent frameworks

For more complex automation requirements where an AI system needs to reason across multiple steps, use tools, and make decisions based on intermediate results, agent frameworks including LangChain, LlamaIndex, and Microsoft AutoGen provide the infrastructure for building autonomous AI workflows. These are more complex to implement than no-code platforms but offer substantially greater flexibility for custom use cases.

Vertical-specific AI tools

Alongside horizontal automation platforms, a growing ecosystem of vertical-specific AI tools has emerged that address automation needs within specific functions. For customer service, Intercom, Freshdesk, and Zendesk all offer native AI agent capabilities. For sales, tools like Clay, Apollo, and Outreach incorporate AI personalisation and automation. For finance, tools like Tipalti, Ramp, and Dext automate specific financial workflows. Selecting the right combination of horizontal platforms and vertical tools is one of the key decisions in any AI automation implementation.

The role of the Anthropic API and similar LLM APIs

Large language model APIs from Anthropic, OpenAI, and Google form the reasoning layer that makes AI workflow automation genuinely intelligent rather than merely automated. These APIs can be embedded into custom workflows to handle the tasks that require natural language understanding: classifying inbound messages, drafting responses, extracting information from unstructured documents, generating summaries, and making context-dependent decisions within defined parameters.

11. How to Implement AI Workflow Automation

The gap between the theoretical potential of AI automation and the operational reality for most businesses is primarily an implementation gap rather than a capability gap. The tools exist. The challenge is deploying them effectively in the context of a real business with real processes, real data, and real people who need to trust and work alongside automated systems.

Start with process mapping

Before selecting any technology, map the processes you intend to automate at a granular level. Document every step, every decision point, every input source, and every output destination. Identify where variability exists and what the decision criteria are at each point. Processes that are not clearly understood cannot be reliably automated. The most common cause of failed automation projects is attempting to automate a process that is not well-defined rather than a failure of the technology itself.

Prioritise by impact and feasibility

Not all processes are equal candidates for automation. Prioritise based on a combination of two factors: the volume and frequency of the process (high-volume repetitive processes yield the greatest return on automation investment) and the degree to which the process is rule-based and well-defined (processes with clear inputs, clear decision logic, and clear outputs are more feasible to automate reliably). Start with high-volume, well-defined processes and build toward more complex or variable ones as confidence and capability develop.

Build in human oversight from day one

Effective AI automation is not about removing humans from processes entirely. It is about removing humans from the repetitive, low-judgment steps while keeping them in the loop for decisions that carry significant consequence, that involve exception handling, or that benefit from human relationship context. Every automated workflow should have clearly defined escalation paths and review checkpoints. Automation deployed without oversight creates risk, damages trust when it fails, and is harder to improve because there is no mechanism for catching and learning from errors.

Measure and iterate

Define success metrics for each automated workflow before deployment: processing time, error rate, cost per transaction, exception rate, customer satisfaction score. Monitor these metrics from the first day of operation and use them to drive continuous improvement. AI automation is not a deploy-and-forget investment. The workflows that deliver the best results over time are those that are actively managed and refined based on operational data.

12. Risks, Challenges, and What to Watch Out For

AI workflow automation carries real risks that are frequently underweighted in vendor presentations and enthusiast commentary. Understanding these risks is essential for implementing automation responsibly and sustainably.

Hallucination and factual errors

Large language models can generate plausible-sounding but factually incorrect outputs. In a customer service context, an AI agent that confidently provides incorrect information about a product, a policy, or an order status can create customer relations problems that cost more to resolve than the automation saved. Mitigating this risk requires grounding AI responses in verified, structured knowledge sources rather than relying on the model's general knowledge, and implementing human review for outputs in high-stakes contexts.

Data privacy and compliance

Automated workflows frequently process personal data. UK GDPR and India's Digital Personal Data Protection Act impose obligations on how personal data is collected, processed, stored, and transferred. AI automation systems that pass customer data to third-party LLM APIs, store conversation logs, or combine data from multiple sources must be designed with data minimisation and lawful processing basis in mind from the outset. Retrofitting compliance into an automation system after deployment is significantly more complex than building it in.

Over-automation and loss of human judgment

Automating processes that genuinely require human judgment, empathy, or relationship context produces poor outcomes. Customer complaints involving genuine distress, complex B2B negotiations, sensitive HR situations, and novel exceptions that fall outside the training distribution of any AI system are all examples where automation should hand off to a human rather than attempt to resolve the situation independently. The design principle is human-in-the-loop for high-stakes decisions, automation for high-volume routine tasks.

Dependency and vendor risk

Building critical business processes on top of third-party AI APIs creates dependency on the continued availability, pricing stability, and capability consistency of those services. API pricing changes, model updates that alter output behaviour, and service outages can all disrupt automated workflows in ways that have direct operational consequences. Mitigation strategies include abstraction layers that allow switching between providers, fallback logic for automation failures, and regular review of vendor relationship terms.

13. AI Automation for SMEs: Starting Without an Enterprise Budget

Enterprise organisations with dedicated technology teams and substantial implementation budgets can deploy comprehensive AI automation programmes across multiple functions simultaneously. For SMEs, the approach must be more selective, more phased, and more focused on quick wins that generate visible return before committing to wider investment.

The three-process starting point

The most effective approach for SMEs is to identify three processes that meet the following criteria: they are high-frequency, meaning they happen at least daily; they are currently consuming a disproportionate amount of team time relative to their strategic value; and they have clear, documentable logic that does not require frequent exceptions handling. Automating these three processes first builds internal confidence, generates measurable savings, and develops the organisational capability to expand automation further.

Tool costs and ROI at SME scale

The cost of AI automation tools has dropped substantially. No-code automation platforms including Make and n8n are available at monthly costs ranging from free tiers to a few hundred pounds per month for mid-range usage. LLM API costs for typical SME automation workloads are often in the range of tens to low hundreds of pounds per month. The ROI calculation for a single well-implemented automation that saves five hours of team time per week is positive within weeks, not months, at these cost levels.

Building internal capability

SMEs that develop internal capability to build and maintain AI automations, even at a basic level, are better positioned than those that depend entirely on external implementation. Investing in upskilling one or two team members in no-code automation tools is a relatively low-cost initiative that creates compounding returns as those individuals identify and implement additional automation opportunities across the business.

14. The UK and India Context: Adoption Patterns and Opportunity

AI workflow automation adoption patterns in the UK and India reflect the different economic contexts, regulatory environments, and business structures of the two markets, both of which are relevant to Prabisha Consulting's client base.

UK market context

UK businesses, particularly in professional services, financial services, retail, and healthcare adjacent sectors, are adopting AI automation with a strong focus on compliance and governance. The regulatory environment around data protection, financial services conduct, and employment law means that UK implementations tend to be more carefully scoped and more conservative in their oversight models than those in less regulated markets. The opportunity is significant: UK businesses with high administrative cost structures in sectors including legal, accountancy, property, and healthcare management are among those seeing the largest proportional cost savings from well-implemented automation programmes.

India market context

India's AI automation adoption is accelerating rapidly, driven by a combination of strong domestic AI talent, high smartphone and digital tool penetration, and a business culture that has been receptive to technology-led operational transformation. Indian SMEs in sectors including fintech, eCommerce, logistics, and IT services are deploying AI automation at a pace that is outstripping many equivalent Western markets. The availability of skilled technical talent at competitive costs also makes custom automation implementation more accessible for Indian businesses than for comparable UK counterparts.

Cross-market operations

For businesses operating across both markets, AI automation offers particular value in the coordination overhead that cross-border operations create: time zone-spanning customer service, multi-currency financial reconciliation, cross-market reporting, and the management of distributed teams. Automation that operates continuously without time zone constraints is especially valuable in these contexts.

15. How Prabisha Consulting Supports AI Automation Implementation

At Prabisha Consulting, we work with SMEs and growth-stage businesses across the UK and India to identify, design, and implement AI workflow automation that delivers measurable operational impact. Our approach is grounded in the same principle that guides all of our work: technology should serve business outcomes, not the other way around.

We begin every automation engagement with a process audit that identifies the highest-value automation opportunities within a client's specific operational context. We prioritise based on a combination of potential cost saving, implementation feasibility, and strategic importance, and we build implementation plans that are realistic for the client's current technology infrastructure and team capability.

Our implementation work spans no-code automation platform configuration using tools including n8n and Make, LLM API integration for AI-powered reasoning within workflows, CRM and marketing platform automation, customer service AI deployment, and analytics and reporting automation that gives leadership real-time visibility into operational performance.

We also work closely with clients on the content and marketing workflow automation that underpins scalable digital growth, ensuring that AI-assisted content production is governed by editorial standards that protect brand quality and search performance.

If you are exploring AI automation for your business and want a partner who will focus on practical, measurable outcomes rather than technology for its own sake, we would be glad to start with a conversation about where the opportunity is greatest in your specific operation.

Visit prabisha.com to get in touch.

16. Final Thoughts: Automation as a Strategic Lever, Not a Cost Exercise

The businesses that extract the most value from AI workflow automation in 2026 are not the ones that approach it purely as a cost reduction exercise. They are the ones that recognise automation as a strategic capability that reshapes what is possible at a given team size and cost structure.

When the routine, repetitive, and process-driven work that consumes a disproportionate share of team time is handled by automated systems, the people in those teams are free to focus on the work that genuinely requires human capability: building relationships, making complex judgments, developing strategy, creating original ideas, and solving novel problems. This reallocation of human attention is where the deepest competitive advantage from automation comes from, not just the cost line improvement.

The 40% cost reduction figure is real and achievable for businesses that implement AI automation thoughtfully across the right processes. But the more important number is the multiple increase in what a well-configured, automation-supported team can deliver, which compounds over time in ways that pure cost reduction arithmetic cannot capture.

The window for early-mover advantage in AI automation is narrowing as adoption accelerates. Businesses that build this capability now will operate with structural advantages over those that wait for the technology to mature further. It is already mature enough. The limiting factor in most cases is not the technology. It is the clarity of process, the quality of implementation, and the willingness to invest in building something that compounds.