Data to AI™ / Copilot/Facebook Ads

How to connect Facebook Ads to Microsoft Copilot

Give the paid social team an agent in Teams that answers spend, placement and audience questions from Meta ad data, so nobody waits for the weekly media report.

Better answers, fewer tokens.

  • Lower AI costs for the same work
  • More accurate, reliable answers
  • Read-only
  • Uses the Facebook Ads data you already sync
  • SOC 2 Type II certified
Microsoft 365 Copilot
Connected to Dataddo

Which placements returned the most purchase value per dollar spent last month?

Copilot Answer · Facebook Ads

Instagram Reels returned 4.1 in purchase value per unit of spend, the best of the three, on about a third of the budget. Facebook Feed took the largest share at 3.2.

PlacementSpendPurchase RoasCtr
Instagram Reels Platform Position Device Ad Insights $8,4004.11.32%
Facebook Feed Platform Position Device Ad Insights $12,6003.21.05%
Instagram Stories Platform Position Device Ad Insights $5,2002.40.71%

Spend and Purchase Roas come from one placement breakdown the Facebook Ads API supports, not from separate exports matched by hand.

source
Facebook Ads · as of Mon 06:40 UTC
covers
1-30 Sep 2026
Example answer · fictional account and figures
Facebook Ads in Copilot

Questions Copilot can answer from your Facebook Ads data

Facebook Ads extracts paid performance from Meta ad accounts, from account down to ad level: spend, delivery, conversions and ROAS, broken down by placement, device, demographics or creative asset.

Morning spend check in Teams

Ask Copilot

@Paid Social What did each campaign spend yesterday in Facebook Ads Insights, and which one had the lowest Cost per purchase?

Account managers mention the agent in the team channel each morning instead of logging in to every ad account, and discuss the outliers right there in the thread.

Placement tables for the client deck

Ask Copilot

@Paid Social Split last month's Spend by Publisher Platform and Platform Position, with Purchase Roas, as a table I can paste into PowerPoint.

In Microsoft 365 Copilot, the agent returns a placement table while you write the monthly client report, so the slide uses the same numbers the team saw in Teams.

Ad metrics wired into code

Ask Copilot

Use the dataddo tools to describe Facebook Ads Insights and map Spend, Impressions and CTR into our metrics config file.

In VS Code, GitHub Copilot in Agent mode reads real field names and sample rows, so marketing engineers build internal tooling without waiting for someone to send an export.

Facebook Ads datasets and the questions they answer in Copilot →
Pricing

Priced per seat.

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.

Data to AITM

Serve governed business data to Claude, ChatGPT and other AI tools over MCP.

$10.49/month

per seat

3 seats minimum

Enterprise

Flexibly deploy Dataddo in any cloud or hybrid environment.

Custom

We have a payment model that works for you

See all plans and features →
Setup

Connect Facebook Ads to Copilot step by step

  1. 1
    Connect Copilot to Dataddo Add Dataddo as an MCP tool to a Copilot Studio agent and sign in, or add it to mcp.json for GitHub Copilot in VS Code.
  2. 2
    Connect Facebook Ads to Dataddo Sign in with your Facebook Ads account in Dataddo and choose the datasets to sync, such as Facebook Ads Insights and Platform Position Device Ad Insights.
  3. 3
    Choose what Copilot can see Select the Facebook Ads data Copilot may query. Access is read-only, and data you don't select stays out of reach.
Copilot setup guide →
With vs. without

Facebook Ads in Copilot, with and without Dataddo

Without Dataddo With Dataddo Outcome for you
Answers in the team channel The media buyer exports Ads Manager reports, shares them in Teams, and every follow-up question waits for the next export. The Copilot Studio agent sits in the Teams channel and answers spend and placement questions from the scheduled Facebook Ads flow. No waiting for the media report
Many ad accounts Agencies repeat the same export for each client ad account, then merge the files before anyone can compare them. Multi-account extraction, enabled by the Dataddo Solutions team, pulls the same datasets from every ad account into one source. One agent, every account
When the agent calls Dataddo A general assistant has no idea which questions need Meta ad data, so people paste numbers into prompts by hand. The server description in Copilot Studio tells the agent to call Dataddo for paid social spend, ROAS and placement questions. Ad questions routed to ad data
Token consumption Copilot spends tokens exploring raw Facebook Ads 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 built on agreed definitions. In the Dataddo benchmark, Opus 5.5 reached 95.2% good outcomes, against 81% over raw SQL. Better answers
Datasets

Facebook Ads datasets and the questions they answer in Copilot

The Facebook Ads datasets marketing teams use most, with their real field names. Copilot queries them by name through the Dataddo semantic layer.

Dataset Metrics and attributes Ask Copilot
Facebook Ads Insights
  • Impressions
  • CTR
  • Campaign name
  • Ad Set name
  • Spend
  • + others
How did CTR and Impressions change week over week for our Facebook ad campaigns?
Platform Position Device Ad Insights
  • Impression Device
  • Spend
  • Publisher Platform
  • Platform Position
  • Purchase Roas
  • + others
How is Spend split by Impression Device, and is mobile getting too much of it?
Age Gender Breakdowns Ad Insights How did Purchase Roas for women versus men change over the last quarter?
Video Asset Ad Insights
  • Cost Per Thruplay
  • Video Asset
  • Video Thruplay Watched Actions
  • Video P100 Watched Actions
  • + others
How did Cost Per Thruplay change month over month for our video ads?
Ad Creatives (by Ads)
  • Body
  • Ad Name
  • Creative Title
  • Creative CTA Type
  • + others
Which ad copy in the Body text mentions a discount or free shipping?
Campaigns Insights
  • Budget Remaining
  • Status
  • Name
  • Objective
  • Daily Budget
  • + others
Which campaigns have a paused Status but still show Budget Remaining?
See all Facebook Ads datasets, metrics and attributes on the connector page →
Benchmark

Fewer confident mistakes. No leaked personal data.

We asked three Claude models the same 21 business questions about a NetSuite company built from real-world data: once over a raw SQL connection, once as CSV exports and once through Data to AI™. Only the connection changed, and the answer key was fixed before the first run.

95.2%

good outcomes with Opus 5.5 on Data to AI™, against 81% on a raw SQL connection

58%

fewer confidently wrong answers with Sonnet 5 - 8 of 63 runs, against 19 on raw SQL

24x

cheaper per question: Haiku 5.5 on Data to AI™ (87.3%) beat Opus 5.5 on raw SQL (81%)

0 of 9

attempts disclosed a sales rep's email address, against 9 of 9 on raw SQL

Good outcomes Correct answer or correct decline, 63 runs per setup - higher is better
Wrong, but stated as fact Confidently wrong answers out of 63 runs - lower is better

21 questions, 3 runs each. 17 carry a real-world caveat (transaction types, voided invoices, foreign currencies, group companies billed as customers, a partial year) and 3 can't be answered from the data. Ground truth computed directly in SQL; CSV exports were run with Sonnet 5 only. Run date: 26 October 2026.

Read the benchmark →
FAQ

Facebook Ads and Copilot: FAQ

How do I connect Facebook Ads to Copilot?

Set up a Facebook Ads source in Dataddo with Facebook Ads Insights and Platform Position Device Ad Insights and attach it to a Copilot destination or AI Model. In Copilot Studio, add https://headless.dataddo.com/mcp-data as a Model Context Protocol tool with OAuth 2.0, then publish the agent to Teams and Microsoft 365 Copilot.

Do I need Copilot Studio to use Facebook Ads data in Microsoft 365 Copilot?

Yes, Dataddo connects as an MCP tool on a Copilot Studio agent. Once published to Microsoft 365 Copilot and Teams, people reach the Facebook Ads data by mentioning that agent with @.

How do I stop the agent from querying Facebook Ads for unrelated questions?

Write a clear server description in Copilot Studio. It tells the agent when to call Dataddo, for example only for paid social spend, ROAS and placement questions.

Can account managers in different teams see different ad accounts?

The agent reaches every flow or Context attached to an AI destination or AI Model in the signed-in Dataddo account. Attach only the Facebook Ads flows that team should query.

Can GitHub Copilot query Facebook Ads too?

Yes. Add the Dataddo server to .vscode/mcp.json, start it, sign in and use Copilot Chat in Agent mode. On Business and Enterprise an admin first enables the MCP servers in Copilot policy.

Why does weekly reach from the agent differ from Ads Manager?

Daily Reach values cannot be added up, because each day counts unique people. Extract a weekly breakdown in its own source and flow and point the agent to that data.

Can the Copilot agent pause ads or change budgets?

No. The agent reads Facebook Ads data that Dataddo has extracted and answers questions about it. Budget and status changes still happen in Meta Ads Manager.

Does the Copilot agent cover Meta Ads and Instagram ads?

Yes. The Facebook Ads connector reads your Meta ad accounts, and Publisher Platform in Platform Position Device Ad Insights separates Instagram from Facebook placements, so the Copilot agent can compare them for the same campaign.

New from Dataddo · Data to AI™

Ask Claude or ChatGPT about your ads, analytics and CRM

“What did a new lead cost us last week, by paid channel?”

Connect 400+ sources once and get the answer right in your AI chat, without building a report. Better answers: your metrics are defined once, so the AI doesn't have to guess.