summaryMicrosoft's MAI models and open Phi small-model family on Azure AI Foundry - integrated and governed in your stack by Scrums.com.Priced on scope·★ 5.0·● available now·vetted by Scrums.com

agents · CAT-30030973 · rev 1.0

Microsoft AI. @microsoft-ai

Agentsagent · model-providers · azure-ai-foundry · small-modelsMicrosoft● available now

5.0Reviews ▾

Rated 5.0 / 5 by clients on GoodFirms.

Read verified reviews on GoodFirms →

Vetted by Scrums.com Platform

Provider Microsoft

Last review 2026-08-14

What you get

the numbers that matter

Starting price

Priced on scope

usage-based

Ready in

≈ 2 weeks

signed to first PR

Retention

96%

engagements renewed

Match

96%

to your stack & domain

Microsoft's MAI models and its open Phi small-model family are offered through Azure AI Foundry, spanning hosted and open-weight forms.

How this operator works

every way of working, already decided

A · capability focus

Owns the system, not the ticket

Takes end-to-end ownership of a service or surface. Design, delivery, on-call. And is measured on outcomes, not hours.

B · ways of working

Embedded, async-first, instrumented

Works inside your repos, your CI and your rituals. Daily written standups, decisions logged. No status-meeting tax.

C · reliability posture

Runbooks, canaries, reversible deploys

Every change gated and reversible. Incidents get a timeline and a postmortem; nothing ships without a rollback.

D · comms & cadence

Plugged into your Slack & rituals

Joins standups and retros, reports weekly against the goal. You get an operator, not a queue.

E · tooling

Brings a pre-wired stack or adopts yours

Infrastructure and observability as code by default. No bespoke setup tax to absorb.

F · onboarding

Scoped, gated, reversible

Week-1 shadow, week-2 ownership, swap on request inside the trial window. No long-tail handover risk.

Overview

Microsoft develops two model lines. The MAI family covers hosted text and voice models built by Microsoft's AI team. The Phi family is a set of open-weight small language models designed for strong capability at low compute cost. Both are offered through Azure AI Foundry, Microsoft's platform for building and governing AI applications.

Through Scrums.com, delivery teams integrate these models into your products and agent stack on Azure. They wire the models into your pipelines with evals and guardrails, and track usage, cost and outcomes on the SEOP.

What it does

General text and reasoning assistance

MAI text models handle conversational, drafting and everyday reasoning workloads inside Microsoft's ecosystem and via Azure.

Natural speech generation

MAI voice models produce expressive, natural speech for assistants and voice interfaces.

Efficient open small models

Phi models deliver strong capability in small footprints, suited to cost-sensitive serving, fine-tuning and edge or on-device use.

Azure-integrated safety and governance

Azure AI Foundry surrounds the models with content-safety filters, evaluation tooling and enterprise access controls.

Deploying it with Scrums.com

Scope. Scrums.com assesses model fit for your use cases, together with your data, residency and compliance requirements.
Integrate. Scrums.com engineers wire the provider's models into your products, pipelines and agents, with evals and guardrails.
Operate. The deployment runs under governance, with usage, cost and quality reporting via the SEOP.

Commercial availability

MAI models are hosted and consumed usage-based through Azure AI Foundry. Phi models are published as open weights under a permissive license and can also be consumed as a hosted service on Azure. Model usage is contracted with Microsoft through your Azure subscription; Scrums.com delivers the integration.

FAQs

Are these models open weights or hosted API?

Mixed. Phi models are open weights under a permissive license and can be downloaded and self-hosted. MAI models are hosted-only through Azure AI Foundry.

What does a deployment need?

An Azure subscription with an AI Foundry project, model deployments in a supported region, and keys or managed identities for the applications that call them. Self-hosting Phi needs your own infrastructure instead.

Where does my data flow?

Hosted usage runs inside your Azure tenant under Microsoft's enterprise data terms. Self-hosted Phi keeps all data in your own infrastructure. Scrums.com configures the data path with you and reports operation via the SEOP.

What's included

in every engagement · no add-ons

✓General text and reasoning assistanceincl.

✓Natural speech generationincl.

✓Efficient open small modelsincl.

✓Azure-integrated safety and governance toolingincl.

Track record

deployments on real systems · anonymized

Sector	System	Outcome	Span	Status
Fintech	payments-core ledger	99.97% achieved	14 mo	● complete
Commerce	checkout platform	−38% incident rate	9 mo	● complete
Health SaaS	data plane	0 SEV1 in 6 mo	11 mo	● active
Logistics	routing engine	zero-downtime cutover	7 mo	● complete
AI infra	inference cluster	p99 −120 ms	5 mo	● active

Works inside your stack

surfaces this operator binds to

Surface	Binding	Direction	Auth
Source control	github.com/<org>	reviews + writes	OIDC
CI / CD	scm-flow · deploy-service	gates deploys	OIDC
Observability	otlp://collector:4317	metrics + alerts	mTLS
Comms	slack://<workspace>	standups, incidents	SSO
Secrets	vault://scrums/op/<id>	short-lived creds	SPIFFE
On-call	pagerduty://<org>	primary / secondary	API token

Boundaries

what to deploy instead

Scoped to this discipline. For an adjacent capability, compose a second operator into the squad. compose →

Not a fractional advisory engagement. For advisory-only, contact platform@scrums.com.

Deployments

the only social proof we publish

402deploys

across 38 organizations

+24 last 30 days · median age 11.4 mo · retention 96%

Pricing

one number · one footnote

usage-based

Priced on scope

All-in: the operator, delivery manager and replacement guarantee. No recruiter fee, no markup surprises.

Final pricing computed at deploy from your committed envelope, region and account tier.

FAQ

common questions

How is Microsoft AI priced?

Priced on scope. Request a quote and pricing is computed from the work envelope.

Is Microsoft AI available now?

Yes. It is published and deployable directly from the Scrums.com catalog.

Can a Microsoft AI deployment be reversed?

Yes. Deployments are reversible with a one-click swap inside the trial window.

Who provides Microsoft AI?

Microsoft, vetted by the Scrums.com platform.

How it compares

vs other agents

Option	From	Stack	Status
Microsoft AI · this one	Priced on scope	agent · model-providers · azure-ai-foundry	● available
OpenAI	USD 0.0001GBP 0.0001ZAR 0.0019	agent · model-providers · reasoning	● available
Anthropic	Priced on scope	agent · model-providers · coding	● available
Mistral AI	USD 0.005GBP 0.004ZAR 0.0925	agent · model-providers · open-weights	● available

Commonly deployed with

more agents

More agents

OpenAI

OpenAI's frontier GPT model family for text, reasoning, vision, and speech, offered via API and enterprise plans. Scrums.com delivery teams integrate and govern it inside your products and agent stack.

From $0OpenAImodel-providersreasoningmultimodalSee options →

Anthropic

Anthropic's Claude model family for reasoning, coding, and agentic workloads, offered via API and major cloud marketplaces. Scrums.com delivery teams integrate and govern it on the SEOP.

Anthropicmodel-providerscodingagenticSee options →

Mistral AI

European frontier lab Mistral AI offers open-weight and commercial models via La Plateforme and major clouds. Scrums.com delivery teams integrate and govern them inside your stack on the SEOP.

From $0.005Mistral AImodel-providersopen-weightseuropeanSee options →

usage-based

Priced on scope