LinkedIn posts
Four made for IT sales and marketing teams, each for one buyer. Open any of them to see it in full.
NNorthwind1 / 7
70%
of GenAI pilots never reach production.
Here’s what the other 30% do differently.
Swipe
NNorthwind2 / 7
01
They pick a boring problem.
Claims triage, not “transform the business”.
NNorthwind3 / 7
02 · FIX THE DATA FIRST
Pilots that reached production had one thing in common.
A named owner for the data71%
Data “owned by IT”18%
Share of pilots that reached production · 64 programs
NNorthwind4 / 7
03
They name a business owner.
IT runs the pilot. Someone in the business owns the result.
NNorthwind5 / 7
04 · MEASURE IN WEEKS
If it can’t show value in 90 days, it won’t.
WK 1–2
Baseline
WK 3–10
Live with real users
WK 11–13
Decide
Day 90: scale or stop
NNorthwind6 / 7
05
They change the workflow.
A new model in the old process is just another tab.
NNorthwind7 / 7
Want the 30-day checklist?
Comment “pilot” and we’ll send it over.
PR
Priya Raman
Head of Data & AI, Northwind
Carousel · Why GenAI pilots stall
LinkedIn posts · 7 pages
THE DATA · 2027 IT BUDGETS
Where enterprise IT budgets are going.
AI & data31%
Security22%
Cloud19%
Applications16%
Infrastructure12%
NORTHWIND CIO SURVEY · N = 214Northwind
NNorthwind2 / 3
WHAT’S GROWING, WHAT’S SHRINKING
AI is being paid for by infrastructure.
AI & data+9
Security+4
Cloud+2
Applications−3
Infrastructure−7
Outsourced run−5
Change in share of IT budget, 2026 → 2027, points
NNorthwind3 / 3
So what for your 2027 plan?
1Fund AI from run savings, not new budget.
2Put security in every AI business case.
3Renegotiate run contracts before Q1.
Full survey: link in comments
Data post · 2027 IT budgets
LinkedIn posts · 3 pages
“
Your data platform isn’t the product. Decisions are.
PR
Priya Raman
Head of Data & AI, Northwind
PRPriya RamanHead of Data & AI, Northwind2 / 4
THE TRAP
We measure platforms by what they hold.
Tables loaded14,000
Dashboards built2,300
Decisions changed?
PRPriya RamanHead of Data & AI, Northwind3 / 4
START FROM THE DECISION
Then work backwards to the data.
1
Name the decision
Which claims to fast-track
2
Find who makes it
The adjuster, 40 times a day
3
Put the answer there
In the claims screen, not a dashboard
4
Measure the change
Cycle time, not usage
PRPriya RamanHead of Data & AI, Northwind4 / 4
What decision would you start with?
Tell me in the comments. I’ll share how three insurers picked theirs.
#DataStrategy#AI#Insurance
Point of view · Data & AI leader
LinkedIn posts · 4 pages
PRPriya RamanHead of Data & AI, Northwind1 / 5
WHAT JUST HAPPENED
Open weights just closed the gap.
Here’s what it changes for regulated banks, and what it doesn’t.
Swipe
PRPriya RamanHead of Data & AI, Northwind2 / 5
THE GAP, IN ONE CHART
From 18 points behind to 3, in two years.
Average score on reasoning benchmarks, best open vs best closed model
PRPriya RamanHead of Data & AI, Northwind3 / 5
WHAT CHANGES FOR BANKS
It’s no longer a quality call. It’s a control call.
OPENCLOSED
DataNever leaves the bankLeaves, under contract
Cost at scaleFalls with volumeRises with volume
Latest skillsMonths behindFirst, always
PRPriya RamanHead of Data & AI, Northwind4 / 5
OUR TAKE
Most banks will run both, behind one gateway.
Customer data in the prompt→ Open, on your cloud
Hardest reasoning today→ Closed, via API
Everything else→ Route by cost and risk
PRPriya RamanHead of Data & AI, Northwind5 / 5
Want our model gateway checklist?
We’re walking bank CIOs through it on Nov 20. Comment “weights” for the invite.
Comment “weights”
Trend post · Open vs closed weights
LinkedIn posts · 5 pages
←
→
1 / 0
LinkedIn posts4 EXAMPLES
Carousel · Why GenAI pilots stall