Why your AI keeps starting from scratch at work
Business AI has plenty of information. The harder problem is maintaining an understanding of what matters as the work changes.
I work on the systems behind B2B SaaS growth.
I’m Karri Takki, a B2B SaaS marketing and GTM operator based in Helsinki.
Currently Founding Growth Marketing Lead at Optivian, where I work close to some of the more interesting questions around AI, sales systems and how business context is maintained over time.
My work tends to cross the usual boundaries
I started closer to sales, moved into growth and marketing, and over time CRM, automation, data and AI became a bigger part of the work.
That progression was not particularly deliberate. It happened because solving a marketing problem often meant fixing something outside marketing.
A campaign can work while the lead process fails. A dashboard can be accurate while the CRM behind it is stale. An AI workflow can produce impressive output while having no reliable understanding of the customer or deal it is supposed to help with.
Those gaps between functions are usually where I find the most interesting problems.
Notes
I mostly write about problems I have encountered while doing the work: CRM adoption, GTM systems, RevOps and what changes when AI starts becoming part of the operating model rather than another standalone tool.
Build vs. buy is often a scoping fight
Internal AI builds often look cheaper because they are solving a much smaller version of the problem.
4 CRM adoption metrics that show whether your pipeline reflects reality
CRM adoption should not be measured only by logins and field completion. The more important question is how closely the system follows what is actually happening.
Work
I have worked across sales, growth and marketing in B2B technology companies, with an increasing focus on the systems connecting them.
Today I’m at Optivian. Before that I led marketing at Futures Platform and Statzon, including work around GTM, positioning, CRM, automation and the launch of Synapse.
What I’m thinking about now
AI is making it much easier to build software, connect systems and automate work.
I’m particularly interested in what happens after the demo: context, data quality, ownership, maintenance and whether the system actually changes how people work.
A lot of my recent notes come back to the same question:
As building gets easier, what becomes the hard part?