One thing is for sure in the age of AI: we’re shipping faster than ever.
Abu has seen it happen at @Snowflake and we see it happening with our projects too.
But speed only counts if the output holds up and that comes down to the data underneath it and the context around it. Clean, governed, connected data gives AI the facts and an ontology tells it how those facts fit together.
That’s the difference between AI that ships something useful and AI that just ships something.
There is no AI stratergy without a data stratergy.
#AfricaTech #AIStrategy #Snowflake #ContextLayer #Assembly
@Revolut just cut 90% of its token spend by switching from @OpenAI to open source / open weight models. Same result, an order of magnitude cheaper. With more control, they now have the ability to fine-tune their own models.
Abu Bakr Khan, Go-to-Market AI Leader at @Snowflake , watched another Snowflake client do exactly that. They swapped the model under their agents twice in three weeks, chasing whichever one was quicker and cheaper that week.
His point to Dimitri and Nic: nobody knows which model wins next year either. Build the data platform underneath to be model agnostic, carrying the access control and governance, and let the model on top keep changing.
The model is the the part of the tech stack you swap. The data platform is the part you build to last.
Episode 7 of The Assembly Line is out now.
#AfricaTech #EnterpriseAI #DataPlatform #AssemblyLine
Salesforce, ServiceNow, Stripe and Oracle agree to disagree on what one client is worth.
None of them are wrong.
None of them have the whole picture either.
That is the gap revenue assurance lives in: not fraud, not a broken system, just four correct answers that do not add up to one. Assembly finds exactly that gap, proven line by line against what is actually delivered.
Recover the revenue you are already owed. Find out more on our website, link in the first reply.
#AssemblyByPygio #RevenueAssurance #Snowflake #EnterpriseAI
Every AI pilot looks brilliant in the demo. Then it hits a thousand use-cases and the token bill arrives.
We sat down with Abu-Bakr Khan, GTM AI Leader for the Middle East, Turkey and Africa at @_snowflake_inc, to talk about what happens beyond the pilot. His view: scaling AI is an economics and governance problem, not a technology one. Models will keep improving. The problem of cost control and data governance between your source systems of action has to be addressed for a successful AI Strategy to hold.
There is no AI strategy without a data strategy.
Episode 7 of The Assembly Line is live. Link in first reply.
#AssemblyLinePodcast #Snowflake #AgenticAI #DataSovereignty
October’s Town Hall welcomed Annie, Kyriakos and 389 new agents to the team!
This TH ran on a report built straight from our own instance of Assembly.
It also advised the team to please all turn our cameras on.
#LifeAtPygio#Snowflake#Assembly#AIEnablement
Every company has the same models. The map of your business is the one thing nobody else has.
Assembly resolves your customers across CRM, ERP and billing, links every decision to its source, and builds that map inside your @Snowflake.
Your business is the one thing competitors can’t copy. Put it to work.
#AssemblyByPygio #Ontology #Snowflake #EnterpriseAI
The model was never the problem. Context is.
Assembly is our answer. A context layer, native in @Snowflake , that turns every system you run into one trusted, traced answer.
We’ve built enough data foundations to know what’s missing.
Find out more, link in the first reply.
#AssemblyByPygio #ContextLayer #Snowflake #DataFoundations
Two of the three systems agree. That is what makes this so hard to catch.
Ten services sold, ten provisioned, seven invoiced. Nothing in the chain is broken enough to raise a flag. The CRM did its job. Service delivery did its job. The billing system invoiced everything it was told about, which was seven.
The uncomfortable part is not the gap. It is that nobody in the business can say which three are missing, because no single record links what was delivered to what was billed.
#RevenueLeakage #BillingIntegrity #DataReconciliation #AfricaTech
Your CFO has to pick one.
One question, three systems, three answers, all of them defensible. The number that ends up in the pack is usually chosen by whoever is assembling it, on the morning it is due, for reasons that are never recorded.
It is not a rounding issue. It is that no centrally agreed definition exists for what counts as “active customer”, as there is no persistent source of truth. Every quarter a few days disappear into reconciling these figures traditionally, and every quarter it happens again.
The reconciling is not the fix (band-aid on a gunshot wound). Agreeing on the definition is. AI needs this structure to work.
#DataQuality #SingleSourceOfTruth #Reporting #AfricaTech
The instinct is always to look for the faulty system. It is the right instinct and it almost never finds anything.
When three systems report different customer counts, each one is usually correct. Finance excludes accounts in arrears. Sales counts signed contracts that are not live yet. Operations counts anything with an active connection. Three reasonable definitions, none of them written down.
So there is no error to find and nobody to correct. The failure sits in the space between the systems, which has no owner and appears in no one's targets. That is why it survives for years inside businesses that are otherwise very well run.
#DataGovernance #EntityResolution #MasterData #AfricaTech
AI tooling is now something most teams can't work without. Nobody planned it that way.
When it has an outage you don't fall back to doing the work manually. You wait. And waiting is a decision your team makes dozens of times a week without anyone measuring it.
The marginal utility you get from working with tokens vs without makes the latter completely unviable.
Out of tokens? Take a lunch break.
Episode 6 of The Assembly Line is in the first comment below.
#AfricaTech #EnterpriseAI #AI #AssemblyLine
The business case was airtight. The return still didn't show up.
Darryl wrote many of those business cases. Months of drafting, tooling to estimate the cost, hundreds of application stacks. The numbers always looked solid going in but not always going out.
He is watching the same thing happen with AI. The return was never about the technology. It was and still is about what you chose to move, and whether it should have moved at all.
Episode 6 of The Assembly Line is out now.
#AfricaTech #EnterpriseAI #CloudEconomics #AssemblyLine