Roadmaps that survive the planning cycle
Sequenced by value and feasibility, with budget and governance attached, so investment decisions survive the next planning cycle.
AI Consulting
Gonzalo Iglesias
AI Strategy · Enterprise AI Adoption · Business ROI
From AI ambition to production outcomes
Advisory, discovery and hands-on delivery across the full lifecycle, from classic Machine Learning to Generative AI.
Executive summary
Enterprises lack a route from AI capability to measurable ROI, not the technology
Engage a partner who goes hand-to-hand to define the business return first, then helps build the AI that delivers.
Challenge
AI investment is outpacing AI return
Capability is no longer the barrier. Tools, models and pilots are widely available.
Use cases are chosen for novelty rather than value or for the P&L line they move.
POCs close without a GO / NO-GO on ROI, so funding stops before production.
What is missing
Business acumen applied before the technology choice
The return is defined, sized and agreed before any build begins.
Where value sits in your business drives the AI architecture and integration.
Delivery is designed for scale and adoption, so the return survives production.
Experience
Seven years turning AI into measured outcomes
7+ years in AI enterprise consulting, 10+ in fast-paced technology innovation.
Teams of 20+ across regions, on budgets above $1M.
5+ industries, including solutions for organisations of 70k+ people.
What you get
A costed path with the return stated up front
A scored use case backlog: value, feasibility and cost for each case.
Production-grade build advisory. Delivered to date: +60% discoverability, 8x time-to-insight, 30x throughput.
Enablement, so your team runs the system after the engagement closes.
Flow of partnership: 45-minute calls for discovery, 2 to 3 weeks for opportunity scan, thereafter a proposal for development.
0Years in AI enterprise consulting
0Team members led, across regions
0Programme budgets directed
0People served by delivered solutions
01 · The ROI gap
Capability is no longer the barrier. Tools, models and pilots are everywhere. What stalls is the route from capability to measurable return. The return is defined before the build, the architecture follows where value sits in your business, and delivery stays accountable through production and adoption.
01 · The ROI gap
Enterprise AI stalls after proof of concept, where experimentation must become delivery
Organizations already understand the value of AI. The struggle is applying it in a manner that gives scalable value and guarantees ROI.
Step 01
Discovery
Identifying where AI creates measurable value.
Step 02
Design
Architecting a solution that fits existing systems.
Step 03
Proof of concept
Proving performance on production data.
Most programmes stop here
Step 04
Production
Scaling, governing and driving adoption.
No line to business value
Use cases are selected for novelty rather than for the P&L line they move, so funding stops at the pilot.
Delivery outside enterprise standards
Each pilot is built as a one-off, so none survives review by IT, risk or audit. Productionalizing an AI solution depends on a knowledgeable design that guarantees ROI when scaled.
POCs treated as a burn zone
A POC should not be a test land to burn and experiment. It exists to understand the metrics and check a GO / NO-GO on the ROI of your solution — with cost, tooling and agent governance already in view.
No adoption plan
The model ships, the workflow is unchanged, and usage flatlines within a quarter. The plan was never for a grounded solution.
The challenge
AI investment is outpacing AI return
- Use cases are chosen for novelty rather than for the P&L line they move.
- Each pilot is built as a one-off, so none survives review by IT, risk or audit.
- POCs close without a GO / NO-GO on ROI, so funding stops before production.
- The model ships, the workflow is unchanged, and usage flatlines within a quarter.
What is missing
The return defined before the build
- The return is defined, sized and agreed before any build begins.
- Where value sits in your business drives the AI architecture and integration.
- POCs exist to produce the metrics behind a GO / NO-GO — not to burn budget.
- Delivery is designed for scale and adoption, so the return survives production.
02 · Coverage
Four connected approaches
Delivered across all four, or integrated into the one capability that is missing.
A scored, costed backlog
Structured sessions that surface, size and rank candidate use cases, and eliminate those that will not repay their build cost.
Hands-on build
Agents, retrieval, knowledge graphs and the pipelines that keep them running under production load.
Usage that sustains
Enablement and voice-of-customer work that converts a deployed model into sustained usage — not a launch spike.
02 · Coverage
Three engagement models let clients enter where the programme is stuck
01Advisory
Recommended for: AI direction not yet clear. Board and executive teams that need a defensible position before committing budget.
- Board-level AI roadmap
- Architecture and build-vs-buy review
- Governance and delivery standards
02Discovery
Recommended for: Use cases not yet prioritised. A need for discovery and a path to scale.
- Use case sourcing workshops
- Value and feasibility scoring
- Data and readiness assessment
- Use case high level design
- POC where necessary
03Implementation
Recommended for: A prioritised case that must reach production.
- Solution design and build
- AI in operations and production hardening
- Enablement and champion networks for adoption
Operating model that follows the nature of your business
Local and mid-size teams
Embedded engineer and architect
Binding constraint: senior delivery capacity.
We design the architecture and ship code alongside your team, at innovation-centre pace: short cycles, current tooling, working software in weeks rather than quarters. Pairing, reviews and enablement mean the system and the team run without us once the engagement closes.
Large enterprise
Design authority and programme lead
Binding constraint: orchestration, governance and decision makers.
Design authority across multiple parallel delivery teams — roadmaps, budgets, dependency management and milestone governance, with the evidence, controls and audit trails that withstand risk and compliance review.
03 · Track record
Delivered programmes lifted content discoverability 60% and time-to-insight 8x
0improvement in content discoverability
0faster time-to-insight for Tax teams
0operational efficiency gain from anomaly detection
0production solutions scaled from 5 pilot assets
0consultants served through a knowledge platform
0documents made searchable in a single corpus
Programme scope: teams of 20+, budgets above $1M, delivery across EY Global Consulting.
03 · Track record
Five capability areas have reached production inside a 70,000-person enterprise
Agent design and development
Multi-step agents scoped to defined workflows, with the guardrails, evaluation and human checkpoints required for deployment in a regulated environment.
Agentic retrieval and search reasoning
+60% Retrieval that plans and decomposes a query rather than returning nearest-neighbour passages, raising answer quality on multi-part questions.
Enterprise knowledge at scale
1M+ A document corpus made searchable, addressing over 60% of the pain points users reported against the previous solution.
Context engineering
Ontologies and knowledge graphs that give models the structure and vocabulary of the business, reducing costs, error on domain-specific terms and hallucinations.
Anomaly detection
30x Detection pipelines on operational data that replaced manual review and raised operational throughput.
Credentials
Two programmes, start to production
Professional services · Knowledge & search
Five pilots scaled to production solutions serving 70,000 consultants
Situation
- Institutional knowledge was distributed across 1M+ documents and dozens of systems.
- Work already completed elsewhere in the firm was routinely rebuilt.
- Five pilots existed with no defined route to production.
What we did
- Designed the GenAI knowledge management strategy with global leadership.
- Defined enterprise ontologies to increase agent accuracy.
- Architected agentic retrieval and search reasoning over the corpus.
- Delivered enablement to ~100 people and captured voice of customer.
Outcome
- +60% improvement in content discoverability.
- Faster time-to-insight.
- Over 60% of reported search pain points addressed.
Role: strategic advisor and architect to executive stakeholders, directing a team of 20+ on budgets above $1M.
Oil and gas · MLOps & anomaly detection
Five pilots scaled to 300+ production solutions through AI MLOps
Situation
- Gas pipe anomalies were compounding and creating a bottleneck in the billing phase.
- Field engineers had the anomaly definition clear, but the process was fully manual.
- Five pilots existed with no defined route to production.
What we did
- Trained 5 different AI models with the anomaly definitions.
- Bootstrapped the models into one standardised detection layer.
- Built around MLOps and delivery practice so pilots could scale.
- Web app development for user interaction and evaluation of anomalies.
- Delivered enablement and captured voice of customer.
Outcome
- 8x increase in operational process.
- 5 pilots scaled to 300+ production solutions.
- 20% increase in anomalies detected.
Role: strategic advisor and architect to executive stakeholders, directing a team of 20+ on budgets above $1M.
04 · Why me
Strategy, engineering and delivery sit with one accountable partner
Where numbers count the most is alongside the empathy for moving your business and solutions forward in the AI era.
C-level and executive sponsors
Translating business goals into funded AI investment cases. Strategic pitches, client reviews and steering.
Programme and product leadership
Roadmaps, budgets above $1M, delivery governance and prioritisation across multiple accounts.
Delivery and engineering teams
Architecture reviews and containerised pipelines, hands-on pairing and technical quality gates.
Localised workforce and end users
White-glove enablement sessions delivered to ~100 people, champion networks, and voice-of-customer capture feeding the backlog.
Sectors
Five data-intensive and regulated industries
- Consumer goods — demand signals, product content and commercial analytics.
- Oil and gas — operational data, anomaly detection and field reporting.
- Banking — regulated delivery, model governance and audit evidence.
- Tax — dense regulatory text, semantic search and document intelligence.
- Telco — high-volume operational data and network optimisation.
- And more — professional services, public sector and industrial clients.
Technology depth
Certified across AI, data and cloud engineering
GenAI & LLMsKnowledge graphsOntologies
Semantic searchMachine learningOCR
Anomaly detectionMLOpsPython
SQLPySparkAzure ML
DatabricksDockerLinux
MLflowBlob & ParquetCI/CD for ML
Certifications
- Neo4j Certified Professional, 2024
- Azure Data Scientist Associate, 2023
- Deep Learning Specialization, 2021
- AWS Cloud Practitioner, 2020
- Machine Learning, Stanford, 2019
- BS Telecom Engineering, UPM
Recognition: EY Internal Business Strategy Program (2025) · McKinsey Datathon City Cup (2019)
FAQs
Questions that come up first
01We know AI is valuable, but where is the ROI here?
That question is answered by the opportunity scan, not by a pilot. Candidate use cases are surfaced across your functions, then scored on value, feasibility and build cost — so what reaches the board is a ranked backlog with a number attached to each case, and the cases that will not repay their cost are eliminated before anyone writes code.
02The tool is good. Now what?
A working tool and a production capability are different things. What usually remains is integration into the existing systems, the governance evidence that gets it past IT, risk and audit, and an adoption plan so usage does not flatline a quarter after launch. That is the Implementation engagement.
03How do we standardise tools, agents and policy across the organisation?
Through delivery standards and governance defined once and applied to every case: an architecture and build-vs-buy position, a shared platform layer, cost and agent controls, and evaluation gates that each solution must pass before it ships. This is Advisory work, and it is what stops each pilot from being a one-off.
04Local inference or cloud? And how do we avoid data leakage?
It depends on the sensitivity of the corpus, the latency budget and the cost curve at your volume — a design decision, not a preference. The architecture review covers where inference runs, what leaves your boundary, and the controls and audit trails that make the answer defensible to risk and compliance.
05Do you do the build, or only advise?
Both, depending on the engagement. With local and mid-size teams the work is embedded: architecture plus shipping code alongside your engineers. In large enterprises the role is design authority and programme lead across parallel delivery teams. On implementation engagements the default is solution architecture, technical direction, design review and enablement, with hands-on development performed by your engineering team or a nominated third party unless the time allocation for the solution permits it.
06How does an engagement start?
A 45-minute discovery call with no preparation required: current position, what has stalled, and whether there is a fit. If there is, a 2-to-3-week opportunity scan produces the scored and costed backlog. From week four, the highest-value case is selected and built with your team in the room.
05 · Next steps
Flow of full partnership
Advisory — Discovery — Implementation. Let the nature of your business decide.
01
Discovery call
45 minutes
Current position, what has stalled, and whether there is a fit.
02
Opportunity scan
2 to 3 weeks
Structured workshops across your functions, producing a scored and costed use case backlog.
03
First build
From week 4
The highest-value case is selected and built, with your team in the room.
Get in touch
Start with the 45-minute call
Tell me where the programme is today and what has stalled. If there is no fit, you will hear that on the call rather than in a proposal.
Emailgiglesiasg@hotmail.com
LinkedInlinkedin.com/in/gonzalo-iglesias-g
Based inMiami · working across regions