Data & Analytics Consulting in 2026
Data & Analytics Consulting in 2026
A conversation about Data Consulting and AI with Jeff Skoldberg
Thank you Johnny Winter, Hwei Geok Ng, Ed Pearson and many others for tuning into my live video with Jeff Skoldberg!
Jeff runs his own data consultancy, Green Mountain Data Solutions, which has been a successful venture since 2018. We talk about what nudged him into consulting and dive into working with clients, running a business and how to size a project. We also walk through a very cool data migration from SQL Server to Snowflake he managed with AI.
Many great points and recommendations were brought up. Below is a list of the most important ones.
Most companies haven’t modernized. We assume everyone uses Claude Code, but that’s talk inside a bubble. General Mills reportedly runs demand planning on spreadsheets, and much of the CPG, warehousing and logistics world is the same. Ramp sits at the other end, with governed datasets and finance teams vibe coding their own apps. Jeff’s clients mostly sit on the beginner side, doing zero to one work.
Skipping data engineering is a disaster. Hoyt has seen a company point Claude at messy source tables to avoid the modeling work. Both failed. A consultancy then landed six to seven months of data engineering work because of that. You can’t hand something “kind of close” to a probabilistic machine and expect good results.
Let AI draft the column descriptions, then let a human own them. Jeff had Claude take a first pass at a couple thousand lines of column definitions. A business analyst then put their fingerprint on them and marked the “do not use” columns. Those columns can be dropped before the gold layer. Add example values too, but never counts, since counts change.
AI migrations are shockingly good. Jeff moved a company off a SQL Server BI system with Cortex Code’s migration skill. It maps every source, itemizes what it takes to turn the old server off, and does a verbatim lift and shift. Then it asks whether you want the transformed SQL as a dbt project or as stored procedures. What took weeks of human planning now takes hours.
Sell the same project, but go further. If a three month migration now takes two weeks, Jeff still sells the three months. He fills the extra time with dashboards, user support and CI/CD. CI/CD used to be a barrier of YAML and terse shell commands. Now AI writes the GitHub Action in a couple of rounds.
Consultants earn their keep through knowing what needs to happen first with data before AI gets involved. Agents can go off on a tangent and build infrastructure you didn’t ask for. People in the trenches know what to do first so you aren’t handing off an AI feature to a stakeholder that could take down prod tables. Adoption is also a fear problem. One client wouldn’t try Snowflake CI/CD simply because they didn’t trust it would work.
Triple your time estimate. If Jeff thinks a report takes one day, he quotes three. If it takes five, he quotes 15. Coming in under budget gives him a conversation about using the remaining hours on improvements. In his words, that’s free sales.
Price from your life, not from a salary. Jeff plans for 45 working weeks a year, not 52, and works backward from there. Contractors also have to cover their own benefits, insurance and taxes. His rate evolved over time, starting with a subcontract gig that doubled what he made in house.
Leverage is how you scale a one person shop. Jeff pairs himself with high quality Latin American contractors at half his rate. Jeff does 20% of the work, the second resource does 80%, and the client gets a great deal. Now when he takes a Friday off, three other people are still billing.
Let content feed the consultancy. Jeff gets paid for a monthly article and webinar for a tech company. He often implements the tool at a client first, then anonymizes the repo for the article. Hoyt calls this cascading content: code, then article, then video. Both agree the best way to learn anything is to write an article about it.
Hi, my name is Hoyt. I’ve spent different lives in Marketing, Data Science and Data Product Management. Other than this Substack, I am the founder of Early Signal. I help data tech startups build authentic connections with technical audiences through bespoke technical content and intentional distribution. Are you an early stage start up or solopreneur wanting to get creative with your technical content and distribution strategy? Let’s talk!
The Full Data Stack is free today. But if you enjoyed this post, you can tell The Full Data Stack that their writing is valuable by pledging a future subscription. You won't be charged unless they enable payments.
© 2026 Hoyt Emerson
548 Market Street PMB 72296, San Francisco, CA 94104