USA flag

U.S.-Based

Azure OpenAI development services for your Azure tenant

We build GPT apps, document search and agents on Azure OpenAI in Microsoft Foundry. Your identity, network and data rules apply from day one.

  • $4,000

    Starting price of a GPT app or integration in your tenant

  • 40

    AI projects delivered since 2020

  • 5.0

    Clutch rating from 40 reviews

Azure OpenAI logo
net-developers-iowa certificate
angular-developers-oklahoma-city certificate
app-development-kansas certificate
ai-company-kansas certificate
ai-company-south-dakota certificate
computer-vision-denver certificate
drupal-developers-wyoming certificate
flutter-developers-maryland certificate
generative-ai-boston certificate
generative-ai-seattle certificate
java-developers-alabama certificate
java-developers-idaho certificate
laravel-developers-denver certificate
machine-learning-kansas-city certificate
nextjs-developer-portland certificate
nodejs-developers-albuquerque certificate
php-developers-little-rock certificate
python-django-developers-denver certificate
react-native-developer-indiana certificate
react-native-developer-nashville certificate
software-developers-albuquerque certificate
software-developers-west-virginia certificate
swift-company-alabama certificate
web-developers-little-rock certificate

How an Azure OpenAI app runs

We approach AI and machine learning as engineering disciplines, not experiments. Our process is designed to move from clear business goals to production-ready models that deliver measurable results.

An analyst asks a question

They ask about a contract in a Teams or web app.

GPT searches your files

Azure OpenAI reads the matching SharePoint documents in your tenant.

The analyst checks sources

Each answer links to the file and passage it used.

The summary is saved

The note is written to the matter or client record.

Azure OpenAI development, in short

  • Kansas City metro software and AI company that builds on Azure OpenAI in Microsoft Foundry
  • GPT apps, document search and agents for Microsoft-based mid-market companies
  • AI inside your own Azure tenant
  • Builds from $4,000, model usage billed at cost through your Azure account
Azure OpenAI builds at a glance

Built in your tenant. Billed through Azure.

  1. 01Build priceFrom $4,000: GPT apps and integrations in your tenant
  2. 02Model usageBilled at cost through your Azure subscription
  3. 03Copilot for your staffRolling Copilot out to employees is a separate service from building on Azure OpenAI
  4. 04You keepCode, prompts and agent setups in your repositories and tenant

What is the Azure OpenAI Service?

The Azure OpenAI Service runs OpenAI models, such as GPT, inside Microsoft's Azure environment under your Azure agreement. Microsoft now offers these models in Microsoft Foundry, where your identity, network and data rules apply to every call.

What Azure OpenAI development services include

  • App and sign-in

    A GPT app, search tool or agent built on Azure OpenAI models in your tenant, with sign-in through Microsoft Entra ID and access that follows your existing roles.

  • Retrieval and guardrails

    Retrieval over SharePoint and other file stores, with sources shown in each answer. Guardrails, which Microsoft previously called content filters, set and tested for your use case.

  • Logging and code

    Logging, evaluations and usage alerts, switched on before release. Code, prompts and deployment settings kept in your own repository.

  • Your endpoint, your bill

    Your apps call the models through an endpoint in your own Azure resource. Usage is billed through your Azure subscription, next to the rest of your cloud spend. Microsoft states that these models don't interact with services that OpenAI operates, such as ChatGPT or the OpenAI API.

  • Azure OpenAI vs Microsoft Copilot

    They are different products. Microsoft Copilot, formerly Microsoft 365 Copilot, is a licensed assistant inside Microsoft 365 apps such as Word, Outlook and Teams. Azure OpenAI is a developer service for custom apps and agents. For a staff rollout of Copilot, see our Microsoft Copilot consulting service.

Azure OpenAI deployment types: where data is processed and how you pay

When you deploy an Azure OpenAI model, you pick a deployment type. It sets where prompts are processed and how Microsoft bills you, so we choose it before the build. For every type, Microsoft keeps data at rest in your chosen geography (Microsoft Learn, deployment types in Microsoft Foundry Models).

Global Standard

Prompts are processed in any Azure region. Fits most workloads: Microsoft's recommended starting point.

Data Zone Standard

Prompts stay inside the US, EU or Asia Pacific data zone. Fits when prompts must stay inside one data zone.

Standard

Prompts stay inside your chosen Azure geography. Fits when prompts must stay in one geography, at lower volume.

Provisioned: global, data zone or regional

Processing matches the scope you choose. Fits steady, high volume that needs predictable speed.

Batch: global or data zone

Any region, or inside a data zone, with a 24-hour target. Fits large jobs that can wait, such as processing a document backlog.

Developer

Any Azure region, with no SLA. Fits testing a fine-tuned model for 24 hours.

Systems we connect to Azure OpenAI

Most Azure builds start with Microsoft 365 data, then reach the line-of-business systems below.

SharePointOneDriveMicrosoft TeamsOutlookDynamics 365Microsoft Entra IDSharePointOneDriveMicrosoft TeamsOutlookDynamics 365Microsoft Entra ID
SalesforceServiceNowSQL ServerSnowflakeYour own APIsSalesforceServiceNowSQL ServerSnowflakeYour own APIsSalesforceServiceNowSQL ServerSnowflakeYour own APIsSalesforceServiceNowSQL ServerSnowflakeYour own APIs
SharePointOneDriveMicrosoft TeamsOutlookDynamics 365Microsoft Entra IDSharePointOneDriveMicrosoft TeamsOutlookDynamics 365Microsoft Entra ID

When Azure OpenAI pays off

Azure OpenAI pays off when your data and users already live in Azure and Microsoft 365. It fits a task that repeats often enough for a custom app to beat a general assistant.

Signs we see before a Microsoft-based company asks for a build:

  • Your security team requires AI to run inside your Azure tenant, under your Azure agreement.
  • Staff search SharePoint, OneDrive or shared drives for the same answers several times a day.
  • A task follows rules a person can explain, such as sorting requests or drafting standard replies.
  • Your product team wants GPT features, and procurement has already approved Azure.
  • An Azure OpenAI pilot works in a demo but has no logging, permissions or usage alerts.

When to wait

If staff mainly need help writing in Word and Outlook, a Microsoft Copilot rollout is usually faster than a custom build. If your company does not run on Azure, the OpenAI API or the Claude API may fit better.

What to prepare before the first call

The Azure subscription and region you plan to use and any data residency rules, read access to the documents or system the app will use or a sample export, a set of real examples (such as 50 past questions, tickets or documents), and one owner on your side who can say what a good answer looks like.

What we build on the Azure OpenAI Service

Five builds Microsoft-based companies ask for most, each deployed in your own Foundry resource.

  • GPT apps in your tenant

    Chat, drafting and summary tools on Azure OpenAI models, with sign-in through Microsoft Entra ID. For example, a drafting tool in Teams that turns meeting notes into a client follow-up.

  • Search over your documents

    Retrieval over SharePoint and other file stores, so answers come from your data and show their source. For example, an analyst asks about a contract clause and gets the answer with a link to the SharePoint file.

  • Agents in Microsoft Foundry

    Prompt agents built in Foundry, or hosted agents that run our code, such as LangGraph, on a managed endpoint. For example, a prompt agent that reads an order status through your API and drafts a reply for support to approve.

  • Assistants API to Responses API

    Foundry's current agent API is the Responses API. We move older Assistants API code and test it before cutover. For example, an assistant with threads and files rebuilt on the Responses API, with answers compared side by side before cutover.

  • Azure OpenAI to Foundry upgrades

    Microsoft lets you upgrade an Azure OpenAI resource to a Foundry resource and keep its endpoint and keys. We plan and test the move. For example, an Azure OpenAI resource moved to a Foundry resource, keeping its endpoint, API keys and existing state.

What you get when the Azure OpenAI build ships

You get a working app in your Azure tenant and everything needed to run it without us: the code, prompts, tests, settings and documentation.

Code and settingsThe app or agent code, merged into your repository through normal code review, with prompts, model versions and deployment settings under version control.
Test set and resultsThe test set and its results, so later changes are checked the same way.
Security switched onEntra sign-in, role-based access, guardrails and network rules in production, plus usage alerts and logging, so finance and IT see every call.
RunbookHow to change a prompt, switch a model version or roll back a release. Nothing runs on a SoftDoes account; maintenance starts at $500 a month if you want us to keep running it.

Azure OpenAI ROI: a worked example

An Azure OpenAI app pays back when the hours it saves cost more than the build and the usage. This example uses assumed numbers for one Azure OpenAI app, not a client result: at these numbers, the build pays for itself in the first month, before Azure usage. Azure bills usage at cost on your subscription, and it depends on volume, model and deployment type. Run your own numbers in the AI ROI calculator, and estimate usage with the LLM cost calculator. After release we track answer quality and usage, so you can compare the real saving with this estimate.

LineAssumed value
Policy and contract questions a month1,500
Minutes to find each answer today10, or 250 hours a month
Minutes with search over your documents4, or 100 hours a month
Loaded cost of an analyst hour$50
Saving a month150 hours, or $7,500
Build cost of a chatbot on your documentsFrom $4,000, once

How Azure OpenAI consulting works

Four steps from the first call to a governed production app.

01
AssessBefore any code

We review your tenant, data, network rules and the models available in your region. You get a build plan and a fixed price.

02
PrototypeIn a test project

We build the app in a test project and measure answers on your documents. You get a working prototype on your data.

03
Build and secureBefore release

We apply Entra identity, role-based access, content filters and network isolation. You get a secured app in your tenant.

04
Launch and monitorIn production

We release to production with tracing, evaluations and usage alerts. You get a production release and a runbook.

AI work in production

Two AI builds from our case studies, both in production, from the team that would build your Azure OpenAI app.

  • Healthcare
2026

Prior Authorization AI

Prior Authorization AI is a healthcare automation startup building AI-powered tools to streamline prior authorization for Medicare and Medicaid medical transportation providers in New Jersey.
Outcome
Replaced a 16-hour manual document collection process with an automated intake and submission pipeline, freeing the founder to focus on clinical oversight instead of clerical work.
  • 16Weekly hours saved
  • 63%Admin time reduced
  • 43%Fewer submission errors
View Full Case Study
Prior Authorization AI case study screenshot
  • Startup
  • Real Estate
2026

Our Haus AI

A finance professional turned founder built Our Haus AI, an AI-powered collaborative workspace where homeowners and contractors scope, bid, contract, and document renovation projects together. SoftDoes took her prototype and made it production-ready.
Outcome
SoftDoes hardened and secured a far-more-than-MVP platform for launch, delivering on a fast, defined timeline so a solo non-technical founder could move to production and start onboarding founding users.
  • 10XMVP scope delivered
  • 2+years Ongoing partnership
  • 1Fully owned Platform
View Full Case Study
Our Haus AI case study screenshot
  • Real Estate
2026

The building buyer

The Building Buyer is a Florida-based real estate investment firm co-founded by Dylan Troiano and Charles Hanlin. They acquire single-family, multifamily, and commercial properties across South Florida and beyond, with a focus on motivated seller opportunities.
Outcome
SoftDoes built a proprietary lead generation and data extraction system that replaced hours of manual research each day, freeing the team to focus on deals instead of data.
  • 18Weekly hours saved
  • 2+Years Ongoing partnership
  • 1Fully owned Platform
View Full Case Study
The building buyer case study screenshot
  • Healthcare
2026

CareInTouch

CareInTouch is a Bay Area home health agency providing skilled nursing and therapy services to patients referred from hospitals and clinics. With over 100 staff, they operate in one of the most complex and compliance-driven healthcare markets in the country.
Outcome
SoftDoes built a custom compliance and billing monitoring app that eliminated missed deadlines, reduced revenue loss, and gave ownership real-time oversight of the entire care workflow.
  • 40% Reduction In Billing Errors
  • 100+Staff Operations Managed
  • 0Missed Compliance Deadlines
View Full Case Study
CareInTouch case study screenshot
  • Startup
2026

Sparkle The Cleaning Service

Sparkle The Cleaning Service is a Detroit-area residential cleaning company with 10 years in business, connecting homeowners with vetted, insured independent cleaners through a subscription-based marketplace platform.
Outcome
Launched a two-sided marketplace that removes the middleman from cleaning transactions, letting cleaners earn more while giving customers a transparent, trust-first booking experience.
  • 75% Platform Transparency
  • 2XPlatform Growth
  • 40%Higher User Retention
View Full Case Study
Sparkle The Cleaning Service case study screenshot
  • Healthcare
2026

FMY Orthodontics

FMY Orthodontics, a multi-location practice in West Tennessee, partnered with SoftDoes to replace a spreadsheet-based financial workflow with a custom web platform. The goal was to simplify how staff present treatment financing while allowing patients and families to review and complete decisions remotely.
Outcome
The new platform streamlined internal workflows and removed manual spreadsheet work while giving patients a more flexible, modern experience. Staff spend less time coordinating financing, and families can review and finalize plans from home with ease.
  • 60%Workflow Reduction
  • 75%Remote Adoption
  • 5Locations Aligned
View Full Case Study
FMY Orthodontics case study screenshot
  • Real Estate
2025

Brizacorp

Brizacorp manages residential properties and coordinates the staff who handle painting, cleaning, and repair requests. Its team tracked work across messengers, phone calls, e-mail, and paper planners, with no single place to manage requests or measure performance.
Outcome
SoftDoes built a web and mobile CRM with role-based panels for admins, managers, and employees, bringing requests, staff, and statistics into one system and making daily tasks faster to complete and easier to track.
  • 98%Application Stability
  • 95%Utilization Rate
  • 30%Faster Task Completion
View Full Case Study
Brizacorp case study screenshot

Risks in Azure OpenAI projects, and how we handle them

Each risk gets a control in code or configuration before release, not a promise in a slide.

What can go wrong
Wrong answers
The model states something false about a policy or contract
Data leaving your region
A Global deployment processes prompts in any Azure region
Oversharing
Search returns a file that a user should not open
Cost overrun
Usage grows faster than planned
Model retirement
Microsoft retires the model version you use
Older agent code
Assistants API code falls behind the current agent API
Blocked answers
Guardrails block valid content in your field
How we handle it
Wrong answers
A test set from your real documents, sources shown in each answer, and a person approves risky outputs
Data leaving your region
A Data Zone or Standard deployment where your rules require it, enforced with Azure Policy
Oversharing
Access follows each user's existing roles, checked with a normal user account before release
Cost overrun
Usage alerts and cost limits per app, and batch deployments for work that can wait
Model retirement
Model versions are pinned, and the test set runs again before any switch
Older agent code
We move it to the Responses API and compare answers side by side before cutover
Blocked answers
Guardrail settings tested on your real examples before release

How Azure OpenAI keeps your data safe

Microsoft commitments, plus our rules

Microsoft states that your prompts, completions, embeddings and training data are not available to OpenAI and are not used to train foundation models without your permission. We add your own access, network and logging rules on top of those commitments. On our side, each request sends only the fields a task needs, keys stay in your secrets manager, and every call is logged. For models that run fully inside your own cloud account, see private AI. For company-wide rules on Azure OpenAI and other AI tools, see AI governance and compliance. Health data has its own rules. Our guide to HIPAA-compliant AI explains what changes when Azure OpenAI reads patient records.

  1. Your geography

    Prompts and responses are processed in your chosen geography, unless you pick a Global or Data Zone deployment.

  2. Encrypted storage

    Stored data, such as Responses API history, stays in your geography, encrypted at rest with AES-256, and you can add customer-managed keys.

  3. Your data, your models

    You can delete stored data at any time, and fine-tuned models are available only to you.

  4. Abuse monitoring

    Microsoft runs abuse monitoring, which can include human review of flagged content. Managed customers can apply to modify it.

  5. Identity and network

    Access runs through Microsoft Entra ID and role-based access, with network isolation and Azure Policy available in Foundry.

Azure AI Foundry is now Microsoft Foundry

Microsoft Foundry is the current name for Azure AI Foundry and, before that, Azure AI Studio. It is the Azure platform where you deploy Azure OpenAI models, build agents and manage access, networking and policies in one place, with tracing, monitoring and evaluations built in.

What changed: Azure AI Services are now Foundry Tools, the Assistants API gave way to the Responses API, threads, messages, runs and assistants became conversations, items, responses and agent versions, a hub with separate resources became one Foundry resource with projects, and monthly api-version parameters became stable v1 routes.

What the change means for your apps

Existing Azure OpenAI resources can be upgraded to Foundry resources and keep their endpoint, API keys and state. New agent work should use the Responses API. Hub-based projects still open in the Foundry (classic) portal, but Microsoft says new investment goes to Foundry projects. Foundry lists more than 10,000 models from Microsoft, OpenAI, Anthropic, Meta and others. Source: Microsoft Learn, What is Microsoft Foundry, read November 2026.

Get a build plan

How to choose an Azure OpenAI consulting partner

Ask any Azure OpenAI consulting partner these questions before you sign, and get every answer in writing.

Questions to ask, and what a good answer includes

  • Which deployment type will you use, and why? A choice tied to your data residency rules, volume and budget.
  • Where will the code and settings live? In your repository and your Azure tenant from day one.
  • How will you measure answer quality? A written test set from your documents and a pass mark agreed before the build.
  • How will permissions work? Entra sign-in, with answers limited to what each user can already open.
  • What will it cost to run each month? An estimate of Azure usage by deployment type, billed at cost on your subscription.
  • What happens when Microsoft retires a model? Pinned versions, a rerun of the test set and a planned switch.
  • What paperwork comes first? An NDA and a certificate of insurance before the first call.

What an Azure OpenAI build costs

A fixed price per build, agreed before work starts. We never resell usage: Azure bills it at cost on your own subscription. Connecting 1 or 2 more systems to the Azure app adds $3,000, and 3 or more add $9,000. See AI agent development cost.

  • AI feature inside your software

    One feature in one app

    From $4,000

  • Custom integration or MCP server

    One system

    From $4,000

  • Chatbot on your own documents

    Answers from your files

    From $4,000

  • AI agent, 1 or 2 connected systems

    With approval steps

    From $7,000

  • AI agent, 3 or more connected systems

    Across several systems

    From $13,000

  • Maintenance and support

    After launch

    From $500 a month

  • Project Discovery

    Credited to the build

    $490

Why companies build on Azure OpenAI with SoftDoes

Companies build on Azure OpenAI with SoftDoes because we ship GPT features on both Azure and OpenAI's own API. Everything we build stays in their repositories and their Azure tenant.

Why companies build on Azure OpenAI with SoftDoes
  • 01

    OpenAI Registered Partner

    We build GPT-powered features and AI agents into business software, on OpenAI's API or on Azure.

  • 02

    Vendor documents first

    Our NDA and certificate of insurance reach your procurement team before the first call.

  • 03

    You own the code

    Code, prompts and agent setups live in your repositories and your Azure tenant.

  • 04

    Claude Partner Network member

    We also build on Claude, which Microsoft lists in the Foundry model catalog, so we compare models on your task.

  • 05

    40 AI projects since 2020

    AI work in production since 2020, from document extraction to filing through Medicare's API.

  • 06

    Rated 5.0 on Clutch

    5.0 from 40 client reviews. Clutch named SoftDoes a Top Artificial Intelligence Company and Top Machine Learning Company in Kansas City for 2026.

Azure OpenAI by industry

The same Azure builds, shaped around each industry's records and rules. Each industry page covers the use cases and compliance rules an Azure build must respect.

  • Healthcare

    Intake summaries, records search and prior authorization forms, with HIPAA controls.

  • Financial services

    Document checks, client letters and policy answers, with an audit log for every AI step.

  • Insurance

    Claims intake and document review, with a person approving every payout decision.

  • Legal

    Contract search, intake and first drafts, written into the case management system.

  • Government

    Records search and form intake for public teams, inside your own Azure tenant.

  • Manufacturing

    Maintenance manuals, quality reports and supplier documents answered from one search.

Questions about Azure OpenAI and Azure AI Foundry

Data, naming, migration and cost. Can't find your answer?

What is the difference between Azure OpenAI and OpenAI?

Azure OpenAI runs OpenAI models inside Microsoft's Azure environment, under your Azure agreement. Microsoft says these models do not interact with services that OpenAI operates, such as ChatGPT or the OpenAI API. Prompts and responses are processed in your chosen geography unless you pick a Global or DataZone deployment. Billing goes through your Azure account.

Does Microsoft or OpenAI use our data to train models?

No, according to Microsoft. Your prompts, completions, embeddings and training data are not available to OpenAI or other customers. They are not used to train foundation models without your permission. Stored data stays in your resource's geography, encrypted at rest, with an option for customer-managed keys. Fine-tuned models are available only to you.

Is Azure AI Foundry the same as Microsoft Foundry?

Yes. Microsoft's documentation lists Azure AI Studio and Azure AI Foundry as earlier brand names of Microsoft Foundry. It is the Azure platform where you deploy Azure OpenAI models, build agents, and manage access, networking and policies. Existing Azure OpenAI resources can be upgraded to Foundry resources and keep their endpoints and keys.

Do we need to move off the Assistants API?

Plan for it. Microsoft lists the Responses API as the current agent API in Foundry, replacing the Assistants API. We recommend moving before you add new features. We map your assistants, threads and files to the new model, rebuild the calls and compare answers side by side before cutover.

Can we use models other than OpenAI in Foundry?

Yes. Microsoft Foundry's catalog lists more than 10,000 models from Microsoft, OpenAI, Anthropic, Meta and others. We pick the model for each task by accuracy, cost and where your data must stay. For Claude features outside Azure, see our Claude API integration page.

How much does an Azure OpenAI project cost?

Builds start at $4,000 for an AI feature or a chatbot on your documents, and AI agents with one or two connected systems start at $7,000. Azure bills model usage separately, at cost, on your own subscription. Maintenance starts at $500 a month. Any project can start with Project Discovery: $490, credited to the build.

What are Azure OpenAI services?

Azure OpenAI services give companies OpenAI models, such as GPT, through Microsoft Azure. You deploy a model in your own Azure resource, call it from your apps and pay through your Azure subscription. Development services build the apps around it, such as search over documents, drafting tools and agents, with your identity and data rules applied.

Is Azure OpenAI the same as Copilot?

No. Microsoft Copilot is a licensed assistant inside Microsoft 365 apps such as Word, Outlook and Teams. Azure OpenAI is a developer service for building your own apps and agents on OpenAI models. Many companies use both: Copilot for everyday writing, and Azure OpenAI for custom tools that reach their own systems.

Is Microsoft Azure AI Foundry free?

Microsoft's cost guide says you are billed for the Azure services a Foundry project uses, such as model deployments, search and storage. Models bill per token or through reserved throughput units, and batch jobs cost 50% less than Global Standard. Our build work is separate: a fixed price from $4,000, with usage billed at cost on your subscription.

What is the difference between Azure AI Foundry and Copilot Studio?

Microsoft describes Foundry as a platform service for building agents, from prompt agents to hosted agents that run your own code, with your choice of models. Copilot Studio is a low-code tool that business teams use to build agents with prebuilt connectors. We use Foundry for custom builds, and Copilot Studio when a low-code agent is enough.

Which Azure OpenAI deployment type should we choose?

Microsoft recommends Global Standard for most workloads, because new models arrive there first and it has the lowest price. Choose Data Zone or Standard when prompts must be processed inside the US, the EU or one Azure geography. Choose provisioned throughput for steady high volume, and batch for large jobs that can wait up to 24 hours.

Can Azure OpenAI keep our data in the United States?

Yes, with the right deployment type. Microsoft says a Data Zone deployment in a US resource processes prompts and responses only within the United States. A Standard deployment keeps processing inside your chosen Azure geography. For every deployment type, data at rest, such as uploaded files, stays in the geography you designate.

Can you connect Azure OpenAI to SharePoint and Teams?

Yes. Most of our Azure builds start with Microsoft 365 data. We build retrieval over SharePoint and OneDrive that follows each user's existing permissions, and Microsoft lets Foundry agents be published to Microsoft 365. Every answer links to the file and passage it used, so staff can check it before they act on it.

How long does an Azure OpenAI project take?

It depends on the app, the data and your approval process. We set the scope, price and schedule of the Azure build in a written plan before work starts. Many teams begin with Project Discovery for $490, which is credited to the build. The plan also lists what we need from you, such as Azure access and test documents.

Updated · Reviewed by Orest Andrusyshyn, Founder and CEO

Flag icon

U.S.-Based

This is a no-pressure, 30-minute conversation. We will talk through what you are building, identify risks or unknowns, and outline what it would take to do it right.

Vendor documentation

Ready for procurement and legal review

Sample certificates of insurance for professional liability (E&O), cyber liability and general liability

Certificate of insurance

  • Professional liability (E&O)
  • Cyber liability
  • General liability
Certificates

Let's build together.

Talk with a senior engineer about your product idea, architecture, and what it would take to build it.

Upload File
By submitting this form, you consent to Come Back Agency, LLC contacting you about your inquiry via email or phone. Your information will be handled in accordance with our Privacy Policy.