We keep hearing that AI has changed the role of the data team. But how so?
Does that mean using the AI feature in your BI tool? Does that mean working mainly out of Claude Code or Codex? Are analysts now the administrators of AI agents for analytics questions?
Here, we try to quantify that. Riverboat Consulting runs DataRoles.dev, a free job board for data practitioners. We pull in job postings from the Greenhouse, Ashby, and Lever ATS — what a lot of the tech industry uses. That gives us a great dataset of job descriptions to analyze. We also sprinkle in some proprietary data from which job postings get the most attention from our users.
What makes this timely: the recent release of Jev by TypeSafe AI. Jev is great at classifying things. In my testing, I found Jev to be more accurate, faster, and cheaper than Claude for performing this analysis. It also is much more flexible than a simple regex to see whether "Snowflake" appears in a job description.
This research report will be a monthly series, and I'm sure will look different next month based on the feedback we get to the first issue. We'll update the method page as we go to reflect any changes.
What type of AI responsibilities get mentioned in data job descriptions?
I checked job descriptions for seven different types of AI responsibilities.
The Jev classifier scored each field as either Specific, Generic, or None. The bar chart below breaks out by classification and includes a Total that is Specific + Generic.
Uses AI tools in own work. Whether or not Claude Code, Codex, Cursor, or vendor-specific AI use was mentioned, and for what purpose. Just listing tool names or saying "AI fluency" gets categorized as generic.
Makes data queryable by agents. Mentions of semantic layers, skills, or other steps to enable AI agents to be able to query data. Just saying "AI-ready data infra" gets categorized as generic.
Builds internal agents or AI analytics. Not just making the data available but actual configuration of agents or AI-powered workflows. Using the agent from a tool like Hex, Basedash, Lightdash, etc. counts.
Tests agent answers. Evals, evals, evals.
Builds AI into the product. Customer-facing AI, not typically expected of analytics teams but something I've seen thought leaders predict.
AI governance and policy. Who governs the use of AI for data work?
Unlocks new capabilities with AI. Transcript enrichment or similar LLM-driven work rather than just coding faster with AI.
In the first view below combined across data engineer, analytics engineer, and data analyst postings, we see 31% of postings describe using AI tools for data work. This was the most likely candidate to rank #1 but also surprising to see that it ranked close enough to making data queryable for agents and building internal data agents.
Use of coding agents now regularly appears on data job descriptions
31% of reviewed job postings for data engineers, analytics engineers, and data analysts mentioned use of coding agents such as Claude Code, Codex, and Cursor or other AI tools to complete work assignments.
Share of postings that ask for each kind of AI work, September 2026.
- Specific
- Generic
- Total = Specific + Generic
Data table
| Bucket | Specific | Specific share | Total | Total share |
|---|---|---|---|---|
| Uses AI tools in own work | 155 | 21.6% | 224 | 31.1% |
| Makes data queryable by agents | 106 | 14.7% | 167 | 23.2% |
| Builds internal agents or AI analytics | 133 | 18.5% | 163 | 22.7% |
| Tests agent answers | 39 | 5.4% | 46 | 6.4% |
| Builds AI into the product | 36 | 5.0% | 43 | 6.0% |
| AI governance and policy | 28 | 3.9% | 32 | 4.5% |
| Unlocks new capabilities with AI | 17 | 2.4% | 28 | 3.9% |
Which AI responsibilities vary by position?
Some of the categories are clearly going to be more relevant for certain data positions than others, so breaking that dataset out below.
Note that the analytics engineer sample size is lowest. This is naturally going to show some selection bias in the data. Legacy or old-school companies are much more likely to have data engineers and data analysts.
Across the top three fields from Figure 1, analytics engineer job postings were the most likely to have mentioned AI use for data work, making data queryable for agents, and building the agents.
This supports arguments that analytics engineers are seeing their responsibilities shift now that coding agents can take on more of the data modeling work, even if that is under the analytics engineer's supervision.
Analytics engineer responsibilities show AI shift faster than data engineers and data analysts
43% of analytics engineer job postings described use of AI tools, 37% described making data available for agents, and 32% described contributing to internal data agents — surpassing data engineers and data analysts.
Share of postings in each role family that ask for each kind of AI work (specific or generic), September 2026.
| Data engineer 369 postings | Analytics engineer 121 postings | Data analyst 229 postings | All families 719 postings | |||||
|---|---|---|---|---|---|---|---|---|
| Bucket | Specific | Generic | Specific | Generic | Specific | Generic | Specific | Generic |
| Uses AI tools in own work | 20.6% | 5.4% | 30.6% | 12.4% | 18.3% | 14.8% | 21.6% | 9.6% |
| Makes data queryable by agents | 17.6% | 10.3% | 25.6% | 11.6% | 4.4% | 3.9% | 14.7% | 8.5% |
| Builds internal agents or AI analytics | 20.3% | 3.5% | 26.5% | 5.8% | 11.3% | 4.4% | 18.5% | 4.2% |
| Tests agent answers | 5.1% | 0.3% | 7.4% | 1.7% | 4.8% | 1.8% | 5.4% | 1.0% |
| Builds AI into the product | 8.1% | 1.9% | 3.3% | 0.0% | 0.9% | 0.0% | 5.0% | 1.0% |
| AI governance and policy | 5.1% | 0.8% | 4.1% | 0.0% | 1.8% | 0.4% | 3.9% | 0.6% |
| Unlocks new capabilities with AI | 3.0% | 0.5% | 1.7% | 3.3% | 1.8% | 2.2% | 2.4% | 1.5% |
Which AI terms are mentioned most often in data job descriptions?
Beyond the classifications, we also did a search for specific terms using a regex to detect the presence of terms, then having Jev review to remove cases where it's the employer describing itself.
Data engineer and analytics engineer job postings look much more similar in this cut. Analytics engineers were by far the most likely to mention context layers, and data engineers were more likely to mention RAG.
Only 15% of data job postings reference AI agents
While data practitioners may primarily work through AI agents in the future, only 15% of job postings first seen in September 2026 mentioned them.
Share of postings in each role family that name each AI topic, September 2026.
| Topic | Data engineer 369 postings | Analytics engineer 121 postings | Data analyst 229 postings | All families 719 postings |
|---|---|---|---|---|
| LLMs | 20.9% | 24.8% | 9.2% | 17.8% |
| Agents | 19.0% | 19.0% | 6.6% | 15.0% |
| Context layer | 10.3% | 37.2% | 7.4% | 13.9% |
| AI coding tools | 11.7% | 16.5% | 8.3% | 11.4% |
| RAG | 10.0% | 5.0% | 0.0% | 6.0% |
| MCP | 4.1% | 3.3% | 0.0% | 2.6% |
| Evals | 2.7% | 1.7% | 0.9% | 1.9% |
What tools are most mentioned in data job descriptions?
Using the same methodology as with the AI terms, we checked for a set of data vendors. It's the usual suspects on top: dbt, Snowflake, Airflow, Tableau, and Databricks.
dbt the most referenced tool in data job postings
45% of job postings first seen in September 2026 mentioned dbt as it is still the standard for data transformation.
Share of postings naming each tool, September 2026.
Data table
| Tool | Postings naming it | Share |
|---|---|---|
| dbt | 324 | 45.1% |
| Snowflake | 250 | 34.8% |
| Airflow | 221 | 30.7% |
| Tableau | 166 | 23.1% |
| Databricks | 163 | 22.7% |
| BigQuery | 161 | 22.4% |
| Looker | 142 | 19.8% |
| Power BI | 141 | 19.6% |
| Redshift | 91 | 12.7% |
| Dagster | 63 | 8.8% |
| Fivetran | 50 | 7.0% |
| Sigma | 30 | 4.2% |
| Omni | 23 | 3.2% |
| Hex | 22 | 3.1% |
| Lightdash | 2 | 0.3% |
| Estuary | 1 | 0.1% |
| Basedash | 0 | 0.0% |
DataRoles.dev September 2026 trends
While we don't collect user data for DataRoles.dev, we do track which jobs get the most views and most people clicking to apply on the company site. For September's data we looked at three angles: work model, pay, and role family.
Raw counts would inflate postings from earlier in the period, so we adjusted for days on site with Adjusted Apply Rate. We may tinker further with this, as rates would be expected to decrease the longer a posting is on the site.
Unsurprisingly, remote jobs outpaced hybrid and on-site jobs. This doesn't necessarily indicate applicant preference. It could just be that remote work models remove the location constraint and open the pool up to many more applicants than jobs located in one metro. Hybrid jobs still had a 37% higher apply rate than fully on-site jobs, though.
Jobs with salary listed had almost double the apply rate of ones that didn't list.
And analytics engineer job postings were the most popular role family. There is again the risk of selection bias due to the makeup of companies in the sample. Simply having analytics engineers may correlate with other factors applicants deem desirable.
Remote jobs still get the most applicants
Remote data jobs had more than 3x the apply rate of on-site jobs, counting clicks to apply on the company site per day a job was listed.
Adjusted Apply Rate, indexed within each group: the lowest segment = 100.
Data table
| Segment | Live postings | Index |
|---|---|---|
| Work model: Hybrid | 535 | 137 |
| Work model: On-site | 1,748 | 100 |
| Work model: Remote | 907 | 328 |
| Pay: No pay listed | 2,483 | 100 |
| Pay: Pay listed | 907 | 196 |
| Role family: Analytics engineer | 468 | 191 |
| Role family: Data analyst | 1,119 | 126 |
| Role family: Data engineer | 1,803 | 100 |
Thanks for reading the first issue of our monthly research report. Please follow up with feedback and anything you'd like us to investigate for the next month.
How to cite
Riverboat Consulting (2026). Data Hiring Signals, September 2026. riverboat.ai/research/data-hiring-signals/2026-09/