Why First-Time AI Projects
Frequently Fail
Most enterprises jump straight to AI implementation. They read a case study, get excited, hire an AI vendor, start building. Six months later, they've spent £150K, have an agent that solves a marginal problem, and wonder why ROI isn't materialising.
The problem: AI projects fail not because of technology risk (that's manageable), but because of strategy risk. You didn't properly diagnose where AI adds value. You overestimated impact. You didn't account for change management. You picked the wrong team to pilot with. The technology was fine; the strategy was weak.
The Cost of Premature Scaling
A Structured 4-Week Assessment
A structured discovery process eliminates strategy risk. We conduct a deep-dive diagnostic covering four critical dimensions of your enterprise logic.
Current State Analysis
What's your current state? Which processes are manual, which are broken, and which could be automated today?
AI Opportunity Mapping
Where does AI add value? Which problems can AI solve for your firm that couldn't be solved before?
Technical Architecture
What's your technical architecture? Defining AI models, private infrastructure, and deep integrations.
Business Case & Risk
What's the business case? Full projection of implementation timeline, cost, ROI, and risk mitigation.
Industry standard: 60%
Promised ROI achieved in Year 1.
The Roadmap to
Execution
The output of this 4-week diagnostic is a detailed 90-day roadmap that you understand, approve, and are confident will deliver value.
We ensure your AI strategy is as robust as your existing systems—without the gamble. 89% of our clients achieve promised ROI within the first year because discovery defines the path correctly.
Current State Assessment
Week 1 is interviews and audits. We work with your team to understand the current state, mapping every stakeholder pain point and technical bottleneck.
Stakeholder Interviews
We interview 12-18 stakeholders across finance (CFO, Controller), operations (COO, managers), sales (VP Sales), and technology (CTO, infrastructure). We ask: What processes take the most time? Where do errors happen most? What systems do you use (CRM, ERP, accounting, etc.)?
System Audit
Technical audit of systems (Zoho, SAP, etc.). We review configurations, API capabilities, and data volume. We generate a data quality assessment: records with critical fields and integration health.
Process Documentation
We map your top 5-10 processes: onboarding, invoice processing, lead management. We document cycle time, error rates, and labor costs. Deliverable: Current State Assessment Report (30-40 pages).
AI Opportunity Mapping
Week 2 is creative and strategic. Given the current state, where does AI create the most material value?
Dimension 1: Impact (Value)
Quantified in: revenue upside, cost reduction (e.g., £104K/yr), risk reduction, or experience improvement.
Dimension 2: Feasibility (Tech)
Can we build it? We assess data, technical requirements, and current AI capabilities.
Dimension 3: Confidence
Probability of success. Straightforward automation: 85-90%. Complex logic: 70-80%.
Dimension 4: Timeline
How long to deliver? Ranges from 8-12 weeks for simple automations to 16-24 weeks for complex integrations.
Dimension 5: Cost
Implementation cost (engineer time) + ongoing cost (LLM APIs, infrastructure). Typically £40K-180K.
Dimension 6: Data Needs
Do we have the data history? Missing data increases both timeline and implementation cost.
Dimension 7: Regulatory
Constraints for FCA (audit trails) or GDPR (dataroom residency). Managed via private model deployments.
Architecture &
Technology Design
Week 3 is technical deep-dive. For the top 3-5 opportunities, we design the technical solution.
Model Selection
For each opportunity, which AI model? Claude (reasoning-heavy tasks), GPT-4o (conversation, structured output), Gemini (multimodal, Google ecosystem), open-source (latency-sensitive, privacy-critical). We benchmark each model against your requirements.
Example: for contract analysis, we test Claude and GPT-4o on 20 sample contracts, measure accuracy, measure cost, measure speed. Claude wins (96.2% vs 91.8% accuracy). Cost difference: £0.12 more per contract, but higher accuracy means fewer human reviews. We quantify trade-offs.
Architecture Design
We design the technical system. Example: lead scoring system looks like: Salesforce webhook → triggers agent → Claude reads lead details + company research + historical data → agent returns score (0-100%) → score written back to Salesforce → Salesforce workflow routes leads based on score.
We specify: which APIs, which databases, which integrations, latency requirements (lead scoring should complete in <2 seconds), error handling (if Claude API fails, what's the fallback), monitoring (how do we know if agent is working), auditing (compliance trail).
Data Pipeline Design
For models that need historical data (churn prediction, forecasting), we design data pipelines: which data sources (your CRM, ERP, data warehouse), how often to refresh (daily, hourly, real-time), which data transformations (cleaning, normalisation, feature engineering), where to store (S3, database, vector store).
We assess data quality: if 30% of records have missing fields, do we clean data first? If we do, cost and timeline increase. We give you the choice.
Infrastructure Design
Where does this run? Cloud-hosted (AWS, Azure, Google Cloud) costs less operationally but requires cloud management. Self-hosted (on your infrastructure) gives you more control but requires ops team. We typically recommend cloud-hosted for UK enterprises (AWS UK region for data residency), with managed services where possible (RDS for databases, ECS for containers, CloudWatch for monitoring) to minimise ops burden.
Technical Architecture Document
Typically 40-60 pages. Includes: for each top opportunity, system architecture (flowcharts, component diagrams), technology stack (LLMs, APIs, databases), infrastructure diagram, security specification, and integration specs.
Pricing & Deliverables Summary
Discovery audits are fixed-scope, fixed-price. No sliding scales. No hidden costs.
Startup Package
1 opportunity deep-dive, lightweight documentation, suitable for smaller firms or single-domain focus.
Standard Package
3-5 opportunities, comprehensive documentation, suitable for most mid-market enterprises.
Enterprise Package
6-10 opportunities, detailed enterprise-grade documentation, multiple business units, suitable for larger enterprises.
Inclusions
Includes: all stakeholder interviews, system audits, technology benchmarking, financial modelling, roadmap planning, delivery of all reports and deliverables.
Exclusions
Doesn't include: actual implementation (billed separately), ongoing support (purchased separately).
Payment Terms
50% upfront (to confirm commitment), 50% upon delivery. If you decide not to proceed with us for implementation, no problem — you have a complete roadmap to take elsewhere. (Though 89% of our discovery clients do proceed with us.)
"If you choose option 1 (move forward with us), we transition to implementation mode. We use the 90-day roadmap as our contract; we execute against it, you pay for implementation (typically £140K-250K for 3-5 agents), and we measure against the KPIs and timelines we defined together."
Strategic Clarification
Why is discovery valuable instead of just jumping to implementation?+
How much time does this require from our team?+
What if the roadmap says don't do AI yet?+
Can we see examples of discovery findings for similar companies?+
Where Discovery Leads Next
Ready to Start Your Discovery?
Stop gambling with AI. Start executing against a tested, structured, and ROI-focused plan.