About
Twegs is an independent consultancy. I run it.
Before starting Twegs, I spent four years inside Google Cloud's AI go-to-market teams. Earlier roles took me from enterprise sales at Oracle to building GTM from zero at an AI robotics startup.
I spent four years embedded in Google Cloud's AI GTM teams, working across Gemini, Workspace AI and Agentspace. Before that, I took go-to-market from zero at an AI robotics startup and worked in enterprise sales at Oracle.
I learned quickly that the workflow is rarely the hard part. It fails when it ignores how the buyer decides, how the team is measured or what sales can credibly use. Technical polish cannot rescue a system that nobody trusts enough to run.
Today I work on the product-marketing problem and the machinery behind it. I might rebuild a position, website, launch or proof programme, then connect the sources and AI workflows that keep it useful. The work has a fixed scope, a named owner and an end.
- Google Cloud
- Four years in AI go-to-market across Gemini, Workspace AI and Agentspace
- AI robotics
- Go-to-market built from zero at an AI startup
- Oracle
- Enterprise sales experience before moving into product marketing
- Twegs
- Independent B2B product marketing and systems consultancy
Positions
Working principles.
These are the calls I make while scoping and running the work.
- 01
Follow the buyer journey across the org chart.
Marketing and sales can both report a good quarter while revenue misses. The useful unit of analysis is the path from first touch to closed deal, including the handoffs between teams.
- 02
Start an AI project by looking for tools to remove.
Adoption is too often treated as a procurement exercise. Consolidating overlapping systems and properly connecting what remains is usually the better first move.
- 03
Fix quality before automating volume.
Content and outreach are easy to scale. If targeting, positioning or qualification is weak, automation only increases the amount of poor work in circulation.
- 04
Customer-facing automation needs a named owner.
A person needs the authority to reject output, change the rules and stop the system. A generic review process is not enough.
- 05
Attribution is degrading, and pretending otherwise is expensive.
Consent gaps, dark social and AI-mediated research have broken the model most dashboards still assume. Self-reported attribution plus incrementality tests beat a confident, wrong model.
- 06
The handover is part of the work.
Every engagement ends with named owners, documentation and training. The project shouldn't create a need for permanent outside support.
Bring me the problem the team keeps reopening.
Describe where buyers lose the thread, which teams disagree and what you have already tried. I will tell you whether I can help.