Enterprise AI implementation is one of the most consequential and most frequently misunderstood technology investments a company can make in 2026. Most organisations budget for the build and forget about data preparation, talent, infrastructure, change management, and ongoing maintenance, the elements that determine whether the investment actually delivers value.
This guide gives you the complete, unvarnished picture of enterprise AI implementation cost: realistic ranges by company size and use case, a breakdown of every cost driver, a comparison of in-house vs. external development, cloud vs. on-premise infrastructure, a week-by-week delivery timeline, and ROI benchmarks by use case.
Why Most Enterprise AI Projects Go Over Budget
The organisations that do see strong returns share a common pattern: they start with a well-scoped use case tied to a measurable business outcome, invest in data readiness before any model work begins, and treat change management as a first-class deliverable.
Enterprise AI Implementation Cost: Realistic Ranges For 2026
The cost of AI implementation for business varies more by organisational readiness and project scope than by any other single factor. Here are realistic ranges by company size, followed by a detailed breakdown by implementation type.
Cost By Company Size
| Company size | Pilot / 1st use case | Year 1 total | Ongoing / year |
|---|---|---|---|
| Small business (20–100 employees) | $10K–$80K | $50K–$200K | $15K–$50K/yr |
| Mid-market (100–1,500 employees) | $80K–$200K | $250K–$900K | $60K–$200K/yr |
| Large enterprise (1,500+ employees) | $150K–$400K | $1M–$5M+ | $300K–$1M+/yr |
Year 1 total cost = build cost × 1.8 to 2.2. A $300K implementation project typically costs $540K–$660K in year one when data readiness, infrastructure, change management, and early ongoing operations are included.
Cost by implementation type
The type of AI implementation you’re building determines the enterprise AI project budget more than company size. Here is the full spectrum:
| AI implementation type | Cost range | What it covers |
|---|---|---|
| AI strategy & readiness assessment | $8K–$25K | Non-negotiable first step. Scope, data audit, use-case scoring. |
| Simple automation (chatbot, rule-based) | $10K–$50K | Pre-built models, minimal data, limited integration. |
| Single agentic use case | $40K–$120K | Well-defined workflow automation with 1–2 integrations. |
| Predictive analytics / NLP application | $100K–$500K | Mid-complexity; structured datasets, model training, production infra. |
| Custom fine-tuned AI model | $75K–$300K | Domain-specific model trained on your proprietary data. |
| AI-integrated SaaS platform | $150K–$500K | AI features are embedded in a new or existing product. |
| Enterprise AI platform (multi-dept.) | $500K–$2M+ | Multi-agent, org-wide integration, full governance layer. |
| Deep learning / autonomous system | $1M–$10M+ | Mission-critical; extensive R&D, high-performance compute, compliance. |
| AI infrastructure build-out | $200K–$1M+ | MLOps, data pipelines, vector DBs, monitoring, security. |
| Change management & training | $20K–$150K | Skipping it averages 40% lower adoption. Do not skip it. |
RELATED GUIDE: How Much Does AI-Powered App Development Cost in 2026? →
Building an AI-powered application on top of existing software or data infrastructure? Our dedicated guide covers AI app cost ranges, model development pricing, and what drives the difference between a $75K and $500K+ AI product build.
The 7 Main Cost Drivers In Enterprise AI Implementation
Data Preparation And Management (30–50% of Total Budget)
Data is the single largest and most underestimated enterprise AI integration cost area. Most organisations start with data that is fragmented, inconsistent, or incomplete. Before any model work can begin, data must be sourced, cleaned, standardised, labelled, and secured. This alone consumes 30–50% of total enterprise AI budgets.
| Data cost item | Typical range | What it covers |
|---|---|---|
| Data discovery & audit | $8K–$30K | Assess what you have, what’s missing, what’s unusable, mandatory first step. |
| Data cleaning & standardisation | $15K–$80K | Deduplication, normalisation, and format alignment across systems. |
| Data labelling (custom models) | $10K–$60K | Human-in-the-loop labelling for supervised learning use cases. |
| Data pipeline build | $20K–$100K | Automated ingestion, transformation, and quality monitoring. |
| Ongoing data management | $5K–$30K/month | Storage, access, governance, and quality checks in production. |
Experience exceptional data engineering services with Liquid Technologies, your trusted partner for innovative solutions. Before enterprise AI can work, your data infrastructure needs to be ready. Liquid Technologies’s data engineering team audits, cleans, and builds the pipelines that make AI implementation reliable and scalable.
RECOMMENDED FIRST STEP: Data Strategy Workshop by Liquid Technologies →
Not sure if your data is ready for AI? Our structured Data Strategy Workshop helps you assess your current data ecosystem, identify gaps, and build the roadmap your AI programme needs before development starts.
THE #1 COST PROTECTION STRATEGY
Invest $8K–$30K in a data audit before any model work starts. This single step prevents the $50K–$200K mid-project surprises that are the most common cause of enterprise AI cost overruns.
Computing Infrastructure (Cloud Vs On-Premise)
AI workloads require GPUs or TPUs for training and real-time inference. The infrastructure decision has a major impact on the enterprise AI project budget:
| Dimension | Cloud (recommended for most) | On-premise |
|---|---|---|
| Upfront cost | Low — pay-as-you-go; no hardware capex | High — servers, GPUs, networking, cooling |
| Monthly cost | Variable $2K–$50K+/month; scales with usage | Fixed once built; predictable at scale |
| Scalability | Instant spin capacity in minutes | Slow procurement cycles, physical limits |
| Data sovereignty | Shared responsibility; region-dependent | Full control required for some regulated industries |
| Best for | Most enterprise AI, variable workloads, early deploys | Classified data, government, very high-volume steady-state |
| Recommended | AWS, Azure, and GCP all have managed ML services | Only if compliance mandates it or scale justifies capex |
Unsure whether cloud or on-premise is right for your AI workloads? Our Cloud Assessment Workshop evaluates your current infrastructure and builds a migration or optimisation plan that reduces cost and supports AI at scale.
AI Talent Acquisition And Development
AI expertise remains one of the most expensive components of enterprise AI implementation costs. The right talent decision, hire vs. partner, has more impact on total cost and timeline than almost any other choice:
| Role | Annual cost (USA) | What they own |
|---|---|---|
| ML / AI Engineer | $130K–$200K/yr | Core model building, training, and fine-tuning. |
| Data Engineer | $120K–$180K/yr | Pipeline build, data quality, integration architecture. |
| MLOps Engineer | $140K–$210K/yr | Model deployment, monitoring, and retraining infrastructure. |
| Data Scientist | $110K–$170K/yr | Experimentation, feature engineering, and model evaluation. |
| AI Product Manager | $130K–$190K/yr | Use-case ownership, stakeholder alignment, and roadmap. |
| AI Governance / Compliance | $90K–$140K/yr | Risk, ethics, and regulatory adherence often overlooked in the initial hiring plan. |
| External AI partner (agency) | $150–$300/hr | Faster than hiring; higher flexibility; IP ownership negotiable. |
AI Model Development And Training
Model development costs depend heavily on the level of customisation required. Simple adaptations of existing foundation models can stay below $20,000. Fine-tuned solutions for a specific domain range from $30,000 to $100,000. Fully custom enterprise models frequently exceed $200,000. Training is not a one-time activity; models must be retrained as business data and conditions evolve, typically every 3–6 months.
For most enterprise use cases in 2026, retrieval-augmented generation (RAG) is more cost-effective than fine-tuning, reducing model development cost by 40–70% while delivering equivalent or better accuracy for knowledge-retrieval tasks.
System Integration With Existing Platforms
AI delivers value only when embedded into real workflows. Integration work includes connecting AI systems to CRMs, ERPs, internal platforms, and external APIs. Mid-sized implementations typically spend $20,000–$80,000 on integration. Large enterprises with legacy systems often exceed $150,000.
INDUSTRY-SPECIFIC GUIDE: Healthcare App Development Cost in 2026 →
Healthcare AI carries distinct integration complexity HL7 FHIR, EHR connections, and HIPAA compliance add $40K–$120K to integration costs. Our dedicated guide covers what healthcare enterprise AI actually costs.
Regulatory Compliance and AI Governance
As enterprise AI adoption expands, compliance obligations grow. Data privacy regulations (GDPR, CCPA, state privacy laws), industry-specific requirements (HIPAA, PCI-DSS, financial AI regulation), and ethical AI standards add 10–20% to overall enterprise AI project budgets and persist throughout the system lifecycle.
Ongoing Maintenance, Monitoring, And Optimisation
Enterprise AI is not a one-time investment. After deployment, AI systems require continuous monitoring for accuracy and bias drift, periodic retraining, infrastructure scaling, and security updates. Annual AI maintenance typically equals 15–30% of the original development cost. A $200,000 system requires $30,000–$60,000 per year to keep running reliably.
In-house AI Development vs. External AI Partner: Cost Comparison
One of the most consequential decisions in any enterprise AI programme is whether to build in-house or work with an external implementation partner. Both approaches are valid, but each has a very different cost profile:
| Dimension | In-house development | External AI partner |
|---|---|---|
| Upfront cost | High Salaries $120K–$300K/yr per role + infra | Lower Implementation fee; no recruitment cost |
| Control & IP | Full ownership of model, data, and architecture | Shared or vendor-owned; negotiate IP clauses carefully |
| Speed to value | Slow 6–18 months to hire, build, and deploy | Faster 4–16 weeks with an experienced partner |
| Customisation | Maximum built exactly to your spec | Varies platform AI is rigid; custom partner is flexible |
| Long-term cost | Lower at scale if AI is core to your business model | Higher at scale if vendor charges per usage or per seat |
| Risk | High Hiring risk, build risk, key-person dependency | Lower initially; vendor lock-in risk grows over time |
| Best for | AI is a core product differentiator or competitive moat | First deployment; regulated data; fixed timeline |
For most organisations implementing enterprise AI for the first time, an external partner is the lower-risk, faster-to-value choice. The in-house model makes sense when AI is a core product differentiator and when you have the runway to build internal capability over 18–24 months.
Liquid Technologies AI Services: Artificial Intelligence Development by Liquid Technologies →
Liquid Technologies’ AI team covers strategy, model development, integration, and post-launch optimisation. See how we structure enterprise AI engagements, industries we serve, and typical project scopes.
How Long Does AI Implementation Take? A Week-By-Week Timeline
How long enterprise AI implementation takes depends on data readiness and organisational alignment more than technical complexity. Simple agentic use cases reach production in 6–12 weeks. Mid-complexity implementations take 16–28 weeks. Complex multi-agent systems spanning departments require 6–12 months.
| When | Phase | Owner | What happens |
|---|---|---|---|
| Wks 1–2 | Discovery & alignment | Strategy + leadership | Business case, use-case selection, data audit, and success metrics defined. |
| Wks 3–4 | Data readiness | Data engineering | Quality check, pipeline gaps, and governance framework drafted. |
| Wks 5–8 | Infrastructure setup | DevOps + ML engineering | Cloud environment, MLOps tooling, vector DB, security controls. |
| Wks 9–12 | Model build/integration | AI engineers | Model selection, fine-tuning or RAG, API integrations, testing. |
| Wks 13–16 | Pilot (UAT) | Product + end users | Limited rollout, user testing, performance benchmarking, iteration. |
| Wks 17–20 | Change management | HR + department leads | Training, workflow redesign, adoption tracking, and feedback loops. |
| Wks 21–24 | Production launch | Full team | Scaled deployment, monitoring dashboards, SLA definition, and support. |
| Month 7+ | Optimise & scale | Ongoing | Model retraining, new use cases, ROI measurement, next phase. |
THE BIGGEST TIMELINE RISK
Data readiness. Poor data quality, siloed systems, and undocumented pipelines add 3–6 months when discovered mid-project. A proper data audit in weeks 1–4 is the most effective timeline protection available.
Hidden Costs Of Enterprise AI Implementation (The Complete List)
The visible development cost is typically 40–60% of what an enterprise AI project actually costs. Here is the complete picture:
| Hidden cost category | Range | Notes |
|---|---|---|
| Data readiness & pipeline work | $30K–$200K | Discovered mid-project in 54% of implementations, the biggest surprise cost. |
| Cloud compute (training + inference) | $2K–$50K/month | Scales with model size, call volume, and real-time requirements. |
| MLOps tooling & monitoring | $1K–$15K/month | Model drift detection, retraining pipelines, performance logging. |
| API and model usage fees | $1K–$10K+/month | OpenAI, Anthropic, Cohere, etc. scales sharply with usage. |
| Model retraining & fine-tuning | $10K–$80K/cycle | Every 3–6 months, not optional models degrade without it. |
| Compliance & ethics audit | $10K–$60K | GDPR, CCPA, HIPAA, and financial AI regulation add 10–20% to total cost. |
| Change management & staff training | $20K–$150K | Most underbudgeted item: 40–80 hrs/user for custom AI; drives adoption rate. |
| Internal AI team headcount (post-launch) | $200K–$600K+/yr | Min. 1–3 dedicated roles needed to sustain and scale AI in production. |
| Vendor/platform licences | $5K–$30K/year | Enterprise AI platforms, vector DB licences, analytics tools. |
| Security review & penetration testing | $10K–$60K | AI-specific risk assessment on top of standard security audit. |
| Legal (IP, DPA, vendor agreements) | $5K–$20K | One-time; essential before any external AI partner or platform is engaged. |
| Infrastructure scaling | $20K–$200K | What starts manageable grows fast as adoption increases across the business. |
TCO RULE OF THUMB
Year 1 total cost = build cost × 1.8 to 2.2. Ongoing annual cost = 15–30% of build for maintenance + $200K–$600K+ for internal AI team headcount. Plan for this before the first invoice arrives.
7 Proven Strategies To Reduce Enterprise AI Implementation Cost
| Strategy | Estimated saving | How it works |
|---|---|---|
| Start with a PoC / pilot | 30–50% total project cost | Validates feasibility before full build commitment. |
| Use foundation models + RAG | 40–70% vs. custom training | No model training cost; knowledge retrieved, not baked in. |
| Choose external AI, partner | 20–40% vs. in-house hire | No recruitment cost, no bench time, no key-person dependency. |
| Cloud-first infrastructure | 50–70% vs. on-prem initially | No hardware capex; pay only for what you use during development. |
| Phased rollout (not big bang) | 25–40% year-1 risk exposure | Fund phase 2 from phase 1 ROI; build only what phase 1 needs. |
| Invest in data quality first | Prevents $50K–$200K overrun | Most mid-project surprises come from data problems found too late. |
| AI-as-a-Service for first use case | 80–90% vs. custom build | Off-the-shelf AI for standard use cases (chatbots, scheduling, analytics). |
Use AI-as-a-Service For Your First Use Case
For standard use cases, customer service AI, document processing, meeting summarisation, HR automation, and commercial AI-as-a-Service platforms offer enterprise-grade capability at a fraction of the cost of custom development. What cost $100,000 to build in 2022 is available as a $30–$100/user/month subscription in 2026. Custom development should be reserved for truly unique use cases that off-the-shelf tools cannot address.
READY-MADE ENTERPRISE AI: Finance AI by Liquid Technologies →
Liquid Technologies’s Finance AI automates reconciliation, invoicing, and financial reporting — a ready-to-deploy solution for a common enterprise use case that delivers measurable ROI in months rather than years.
READY-MADE ENTERPRISE AI: HR AI by Liquid Technologies →
Liquid Technologies’s HR AI handles end-to-end automation from recruitment to offboarding, one of the highest-adoption first AI use cases for mid-market companies.
READY-MADE ENTERPRISE AI: Machine AI by Liquid Technologies →
Liquid Technologies’ Machine AI reduces equipment downtime through AI-powered diagnostics and SOPs. A strong first use case for manufacturing, logistics, and engineering-heavy enterprises.
Enterprise AI ROI: How Long It Takes And What Drives It
| Use case | ROI timeline | What drives the return |
|---|---|---|
| Invoice & AP automation | 6–9 months | 68–78% reduction in manual processing; immediate measurable cost saving. |
| IT helpdesk AI agent | 4–8 months | 68% of Tier-1 tickets were resolved without human escalation. |
| AI-assisted software development | 3–6 months | 40–55% more code per week; faster releases, lower QA cost. |
| Customer service AI | 6–12 months | Reduced headcount growth; CSAT improvement from 24/7 resolution. |
| Predictive maintenance (manufacturing) | 8–18 months | 45% downtime reduction; 25% lower maintenance cost. |
| AI analytics/demand forecasting | 12–24 months | Revenue impact from better decisions; harder to isolate in P&L. |
| Multi-agent enterprise platform | 18–36 months | Highest investment; highest ceiling — compounding returns over time. |
Best First Use Cases For Enterprise AI (Fast Roi, Manageable Cost)
Choosing the right first use case determines whether your enterprise AI investment builds momentum or stalls. The best first use cases combine structured data that already exists, a clear measurable baseline, and a defined user group who will adopt the tool.
Proven High-Roi First Deployments
- Invoice and AP automation structured data, clear baseline, 6–9 month payback
- IT helpdesk AI agent 68% Tier-1 ticket resolution without human escalation; 4–8 month ROI
- AI-assisted code generation for your dev team 40–55% productivity gain; 3–6 month ROI
- Customer service AI for common queries, 24/7 resolution, measurable CSAT improvement
- Document processing and summarisation 70–85% time reduction on repetitive review
RELATED GUIDE: How to Integrate AI into an Existing App →
Not starting from scratch? Most enterprise AI implementations sit inside or alongside existing software. This guide covers how to add AI features to an existing platform without a full rebuild.
Use Cases To Avoid As Your First Implementation
- Custom model development before you’ve tested whether RAG or an off-the-shelf approach works
- Anything requiring a complete data infrastructure overhaul before it can function
- Multi-department systems with no clear owner or executive sponsor
- Open-ended ‘AI transformation’ projects with no defined deliverables or success metrics
- Customer-facing generative AI without brand, legal, and compliance sign-off
Enterprise AI Implementation Readiness Checklist
Before committing the budget, run this readiness check. Organisations that skip this step consistently encounter data and governance problems mid-project that cause cost overruns and timeline delays:
| Readiness area | Signal | What it means |
|---|---|---|
| Data quality & access | Ready if: | Core business data is structured, accessible, and accurate in >80% of records. |
| Executive sponsorship | Ready if: | A named C-suite sponsor owns the AI agenda with real budget authority. |
| Technical infrastructure | Ready if: | A cloud environment exists; data pipelines are modern or can be modernised. |
| Defined use case & metrics | Ready if: | You can describe the workflow, baseline metric, and success definition clearly. |
| Change management capacity | Ready if: | HR and department leads are aligned on training and workflow change commitment. |
| AI governance framework | Not ready if: | No AI policy, data privacy controls, or model risk review process exists yet. |
| Compliance requirements | Clarify first if: | Regulated industry (healthcare, finance, legal) compliance architecture must be defined before building. |
If you can’t answer yes to the first five areas, the right first investment is an AI strategy and readiness assessment ($8K–$25K), not a full implementation. It typically saves 30–50% of total project cost by preventing mid-project surprises.
A structured 2-week engagement. We identify your highest-value use case, assess data readiness, score your enterprise AI readiness across all seven dimensions, and deliver a realistic cost and timeline before you commit to a full build.
What a $280,000 Enterprise AI Implementation Looks Like
EXAMPLE PROJECT: A Houston-based professional services firm (280 employees) needed an AI system to automate contract review, flag non-standard clauses, and route documents, replacing a process consuming 3 FTEs and averaging 4.2 business days per contract.
- AI strategy and readiness assessment: 2 weeks, $18,000 data audit, use-case scoping, success metrics
- Data preparation: 4 weeks, $35,000 contract corpus cleaning, metadata tagging, pipeline build
- Model development: 8 weeks, $95,000 fine-tuned LLM on contract corpus, clause classification
- System integration: 4 weeks, $45,000, document management, email routing, CRM connection
- Change management and training: 3 weeks, $28,000 workflow redesign, training, and adoption tracking
- UAT and production deployment: 3 weeks, $22,000 staged rollout, monitoring dashboards
- Year-1 infrastructure and maintenance: $37,000 cloud compute, model retraining, support
Total year-1 cost: $280,000. Timeline: 24 weeks. Outcome at month 12: Review time reduced from 4.2 days to 11 hours. 2.1 FTEs reallocated to higher-value work. Annual cost saving: $340,000. ROI breakeven: 11 months.
ALSO BUILDING CUSTOM SOFTWARE?: Custom Software Development by Liquid Technologies →
Many enterprise AI projects sit inside a broader custom software build. See how Liquid Technologies approaches end-to-end custom development, from discovery and design through deployment and post-launch support.
Why Houston Enterprises Choose Liquid Technologies For AI Implementation
Liquid Technologies is a Houston-based software and AI development firm. We implement enterprise AI across healthcare, logistics, fintech, and professional services, from first use-case scoping through production deployment and ongoing optimisation.
- AI strategy workshops structured 2-week engagements to identify, score, and sequence the right use cases
- Data readiness first, we audit your data infrastructure before any model work starts.
- Senior-led implementation, no handoff to junior teams post-kickoff
- Regulated industries: HIPAA-compliant AI, financial data security, SOC 2-ready architecture
- Change management is not an optional add-on
- Transparent pricing T&M by default, with sprint-based delivery and weekly budget reviews
BUILDING AI INTO A MOBILE APP?: Mobile App Development by Liquid Technologies →
Many enterprise AI implementations include a mobile interface or field app. See how Liquid Technologies builds mobile components that integrate with AI backends for logistics, healthcare, and field-service use cases.
RELATED READING: OpenAI Consulting for Enterprise: How It Works →
Using OpenAI’s GPT models as the foundation for your enterprise AI? Our guide covers how OpenAI consulting engagements work, what they cost, and when they’re the right choice vs. building on open-source models.
Frequently Asked Questions
How much does enterprise AI implementation cost?
Enterprise AI implementation cost ranges from $10,000–$50,000 for a simple automation pilot to $1 million–$10 million+ for a large-scale deep learning or multi-agent enterprise platform. Most mid-market companies (100–1,500 employees) should budget $250,000–$900,000 for a well-executed first year, including data readiness, infrastructure, change management, and early ongoing operations. Year 1 total cost is typically 1.8–2.2× the core implementation estimate.
What is the cost of AI implementation for small businesses vs. large enterprises?
Small businesses (20–100 employees) typically spend $50,000–$200,000 in year one, with pilots starting at $10,000–$80,000. Mid-market companies (100–1,500 employees) spend $250,000–$900,000 in year one. Large enterprises (1,500+ employees) typically budget $1 million–$5 million+ for year one, with ongoing annual costs of $300,000–$1 million+. The biggest cost variable across all sizes is data readiness, which can add $30,000–$200,000 regardless of company size if data quality problems surface mid-project.
What is the AI integration cost for enterprise systems?
AI integration cost for enterprise connecting AI systems to existing CRMs, ERPs, internal platforms, and external APIs typically runs $20,000–$80,000 for mid-sized implementations and $100,000–$150,000+ for large enterprises with legacy systems. Healthcare AI integration (HL7 FHIR, EHR) adds $40,000–$120,000. Projects with five or more integrations should budget 20–30% of the total project cost specifically for integration work. Integration is consistently the second most underestimated cost driver after data preparation.
How long does AI implementation take for an enterprise?
How long enterprise AI implementation takes depends primarily on data readiness and organisational alignment. Simple agentic use cases reach production in 6–12 weeks. Mid-complexity implementations take 16–28 weeks. Complex multi-agent systems spanning departments require 6–12 months. The biggest timeline risk is discovering data quality problems mid-project, which adds 3–6 months. A proper data audit in weeks 1–4 is the most effective timeline protection available.
What are the hidden costs in an enterprise AI project budget?
The most commonly missed costs are data readiness and pipeline work ($30K–$200K, discovered mid-project in 54% of implementations), cloud compute ($2K–$50K/month), model retraining every 3–6 months ($10K–$80K/cycle), change management and training ($20K–$150K most underbudgeted), internal AI team headcount post-launch ($200K–$600K+/year), and AI-specific compliance and security audits ($10K–$60K). Budget year 1 total cost at 1.8–2.2× your core build estimate.
Should I build AI in-house or use an external AI implementation partner?
For most organisations implementing enterprise AI for the first time, an external partner is the lower-risk and faster-to-value choice. In-house development requires hiring ML engineers ($130K–$200K/year), data engineers ($120K–$180K/year), and MLOps specialists, plus 12–18 months before the team is productive. External partners deliver the first deployment in 4–16 weeks with no recruitment cost or key-person dependency. The in-house model makes sense when AI is a core product differentiator, and you have the runway for a long-term capability build.
What is the ROI on enterprise AI implementation?
Enterprise AI ROI varies by use case and execution quality. Simple operational use cases (invoice automation, IT helpdesk, code generation) typically achieve ROI in 4–12 months. Analytics and customer-facing AI take 12–24 months to show P&L impact. McKinsey data shows that only 6% of organisations are currently achieving significant enterprise-wide EBIT impact from AI, but those that do are 3× more likely to have fundamentally redesigned their workflows rather than layering AI on existing processes.
