Data to AI™ / Claude/Gusto
Get headcount, attrition and leave answers from Gusto in Claude, while names, SSNs and home addresses stay hashed inside Dataddo.
Better answers, fewer tokens.
Claude
What did each September payroll cost us, including employer taxes?
The three September payrolls came to $412,300 in Gross Pay and $31,540 in Employer Taxes. The off-cycle run on Sep 19 covered bonuses only.
| Check Date | Gross Pay | Employer Taxes | Net Pay |
|---|---|---|---|
| Sep 15, 2026 Payrolls | $196,800 | $15,060 | $141,900 |
| Sep 19, 2026 Payrolls | $18,700 | $1,430 | $12,600 |
| Sep 30, 2026 Payrolls | $196,800 | $15,050 | $142,100 |
Payrolls reads processed and unprocessed payrolls in the date window, so the answer checks Processed. Gross Pay and Employer Taxes were left out of Hash Sensitive Fields here so they can be summed.
Gusto runs payroll, benefits and HR for small and mid-sized businesses. Dataddo extracts employees, jobs, terminations, payrolls, per-employee pay, time off and contractor payments for one Gusto company, so HR and payroll teams can ask about headcount, payroll cost and leave in plain language.
Count Employees by Department and Current Employment Status, then list Employee Terminations by Effective Date for last month per Department.
The HR lead drafts the monthly people update for the leadership team in Claude, with headcount and leavers per team, without opening a single export.
From Employee Jobs, how many Primary jobs have a Hire Date in each quarter of the last two years, grouped by Title?
Before the hiring plan review, the people partner shows which roles were filled on schedule and where hiring slowed down after the spring freeze.
Use the dataddo server to pull Employee Terminations with Effective Date and Department for 2025 and 2026 and write data/gusto-attrition.csv.
A people analyst runs Claude Code next to the attrition notebook; the CSV holds hashed identifiers only, so it can be committed to the team repository.
Data to AI™ is priced per seat, like the AI assistants your team already uses, with ~25% off on annual billing. Need tailored contracts or volume pricing? Talk to us about Enterprise.
Serve governed business data to Claude, ChatGPT and other AI tools over MCP.
$10.49/month
per seat
3 seats minimum
Flexibly deploy Dataddo in any cloud or hybrid environment.
Custom
We have a payment model that works for you
Gusto datasets with personal data offer the Hash Sensitive Fields option. Dataddo replaces the values of the columns you select with a hash before the data leaves Dataddo, so Claude can still count and group records but never sees the raw values.
| Dataset | Sensitive fields preselected by Dataddo |
|---|---|
| Companies |
|
| Contractors |
|
| Employees |
|
| Employee Terminations |
|
| Employee Work Addresses |
|
| Payrolls |
|
| Without Dataddo |
With Dataddo
|
Outcome for you | |
|---|---|---|---|
| Personal data in the chat | An HR coordinator exports the employee list from Gusto and uploads it, so First Name, SSN and Date of Birth for every person land in the conversation. | Hash Sensitive Fields hashes those columns during extraction, so Claude can count people per Department but never reads who they are. | No raw identities in Claude |
| Leavers per team | Someone filters the terminations report by date, matches each leaver to a team by hand and recalculates the turnover rate each month. | Employee Terminations carries Effective Date and Department, so Claude groups leavers per month and team straight from the extracted data. | Attrition without manual matching |
| Organization access | Every HR team member keeps their own copy of the headcount sheet, and the numbers in leadership decks disagree. | A Claude Team or Enterprise Owner adds the connector once, and each HR member signs in with their own Dataddo account, which limits what they can reach. | Access follows Dataddo accounts |
| Token consumption | Claude spends tokens exploring raw Gusto columns, sampling rows and retrying until it understands the data. | Field definitions and relations are supplied up front, so the model skips the exploration and goes straight to the answer. | Fewer tokens per answer |
| Answer quality | Prone to wrong joins, invented fields and numbers that are hard to check. | Answers grounded in defined fields. In the Dataddo benchmark, 78.5% of questions were answered correctly, against 65.5% with plain CSV files. | Better answers |
The Gusto datasets HR and payroll teams use most, with their real field names. Claude queries them by name through the Dataddo semantic layer.
| Dataset | Metrics and attributes |
Ask Claude
|
|---|---|---|
| Employees |
|
How many employees does each Department have, split by Current Employment Status? |
| Employee Jobs |
|
How many people were hired per quarter over the last two years, based on Hire Date? |
| Employee Terminations |
|
How many terminations took effect per month this year, by Department? |
| Payrolls |
|
What were Gross Pay and Employer Taxes per Check Date over the last three months? |
| Payroll Employee Compensations |
|
How many employees were paid on each Check Date this quarter, counting distinct Employee UUID? |
| Time Off Requests |
|
How many time off requests are waiting with each Status right now, by Policy Type? |
| Employee Time Off Activities |
|
Which employees have the highest remaining Balance for each Policy Name? |
| Contractor Payments |
|
What was the total Payments Wage per Contractor ID over the last six months? |
We asked 43 pre-registered questions of three real datasets - Google Search Console, HubSpot deals and Google Ads - and delivered the same rows to the same model in four ways. Only the data layer changed.
of questions answered correctly - the best of the four delivery methods tested
fewer confidently wrong answers - 14.0% of runs against 24.0% and 26.3%
more correct answers on messy CRM data - better on 8 questions, worse on none
Same model, prompts and rows in every condition; 759 graded runs, ground truth frozen before any run. On Google Search Console alone, documented CSV files did as well as Data to AI™ - the advantage comes from messier data such as HubSpot deals.
Read the benchmark →Create a Gusto source in Dataddo, authorized with OAuth by an account with admin-level permissions, pick the company and datasets such as Employees and Employee Terminations, and turn on Hash Sensitive Fields. Attach it to a Claude destination or AI Model, then add Dataddo Data Access from the Claude connectors directory.
One with admin-level permissions in Gusto, authorized through OAuth. When you create the source, you choose the company to extract. Claude never sees the Gusto credentials.
Not if you hash them. With Hash Sensitive Fields on, fields such as First Name, Last Name, SSN, Date of Birth and Home Address Street 1 in Employees are hashed before the data leaves Dataddo, so Claude only receives hashed values.
In Payrolls, figures such as Gross Pay, Net Pay and Employer Taxes are preselected for hashing. If you want Claude to total payroll cost, remove those columns from Select Columns for Sensitive Computation.
Employees, Employee Jobs and Employee Terminations pull all current records on each scheduled run, so Claude reports the state as of the last extraction.
No. Claude queries extracted data through the semantic layer by entity and field name. Dataddo never accepts SQL from the model and nothing is written back to Gusto.
Yes. Claude shows the connector's tool permissions, and you decide per tool whether it may run freely or must ask you first, which suits sensitive HR questions.
