Helkyn Coello Costa
Fractional CTO / Chief AI Officer taking enterprise B2B software from AI pilots to production — governance, cost control, shipped results
Operating focus
Industry experience
Enterprise B2B SaaS · Accounting & Financial Operations Software · Real Estate Technology · Fintech · Healthtech
Company stages
Growth · Scale-up · Bootstrapped / Profitable SMB · Established Enterprise · PE-backed
Operating regions
United States · Germany & DACH · Europe · Latin America · Remote-distributed teams
Languages
English · Spanish · Italian · German (B1-B2)
Executive statement
I lead enterprise technology and AI platform work for a US B2B software company, and I take fractional and advisory mandates where the question is not whether to use AI, but how to run it safely inside a business that already has paying customers. Twenty-five years of enterprise platform ownership sits behind that: architect and lead developer before I led the organisation, and three complete architectural generations delivered — Access-based legacy to client-server in 2005, client-server to fully web-based between 2017 and 2025, and now AI-native — each one shipped with customers live in production throughout. As Chief AI Officer at AccountTECH I own company-wide AI strategy, platform architecture and execution across an accounting and financial-operations platform serving roughly 300 client companies and 15,000 users. Concretely: a retrieval-augmented knowledge system over client training material, internal documentation and video transcripts indexed to the second; a rules-based anomaly-detection engine reading accounting ledgers for unusual payment amounts, out-of-pattern transactions and probable duplicates; automatic test generation for changes made by non-engineers, currently around 90% of a script produced automatically; and a single company-wide standard for AI-assisted development so contributions arrive consistent, reviewable and safe to ship. I am equally willing to say no. I ran a self-hosted LLM pilot and discontinued it after evaluation — the hardware would have been obsolete within two years — moving to rented infrastructure on a business case that avoided roughly EUR 100k of capital expenditure and an estimated EUR 70k a year in token spend. I evaluated open-source agentic frameworks and declined adoption on enterprise-readiness and security grounds. Before the AI mandate I ran a 15-person engineering and quality organisation, built the quality function from zero, owned DevOps and CI/CD, and introduced Playwright end-to-end automation that cut regression cycles four- to fivefold. I currently lead the SOC 2 programme (Vanta deployed across Google Workspace, GitHub, AWS and Cloudflare) and authored the company's AI governance framework using the EU AI Act as a reference. Most of this work is change management wearing an architecture diagram. I am based in Munich, work in English, Spanish and Italian, and operate comfortably across US, European and Latin American time zones.
Leadership scope
Chief AI Officer with a company-wide mandate across products, engineering and internal operations at a US B2B enterprise software company (approximately 300 client companies, 15,000 users). Previously VP of Development, owning a 15-person engineering and quality organisation through a multi-year re-platform, with direct responsibility for architecture, DevOps and CI/CD, cloud vendor selection including commercial terms, release practice, and resolution of critical production incidents. Decision rights today cover AI platform architecture, model and vendor selection, infrastructure spend and governance policy; accountable to the CEO and executive team for AI strategy and the SOC 2 programme. A large part of the AI mandate is executed through influence rather than line authority, since the delivery teams do not report to me. Regulated-adjacent context throughout: accounting and financial data, SOC 2, EU AI Act. Also set technology direction and managed a small development team as Director of Technology for a medical services company in Lima.
Selected outcomes
- Avoided roughly EUR 100k of capital expenditure by discontinuing a self-hosted LLM pilot and moving to rented inference infrastructure.
- Delivered three complete architectural generations of a live enterprise platform with customers in production throughout.
- Own company-wide AI strategy and platform architecture for an accounting and financial-operations platform serving 15,000 users.
- Built the quality function from zero inside a 15-person engineering organisation, significantly reducing defects reaching production.
- Established the company's AI governance framework (model approval, acceptable use, documentation, logging, human oversight) using the EU AI Act as a reference.
- Leading the SOC 2 certification programme, with Vanta compliance monitoring deployed across Google Workspace, GitHub, AWS and Cloudflare.
Profile verified 2026-08-19. Direct contact details are shared only after a mutual introduction.