Strawberry vs Microsoft Copilot

Most of the job happens in tools your Microsoft subscription does not include.

Think about the request that lands in your inbox and quietly costs you a Thursday.

Say it is the security questionnaire a prospect sent your AE. Sixty questions, due Friday. Every answer already exists, in last quarter's questionnaire, in the policy doc, and in the ticket where single sign-on was finished. All sixty have to be typed into the prospect's portal.

None of it is hard. It is slow because the answers sit in five places and the form sits in a sixth.

Microsoft documents Copilot pairing with the Microsoft 365 apps, and using content in Microsoft Graph to personalize responses with a user's work emails, chats and documents.

Strawberry starts in the sixth. Your AI companion opens the portal you are signed into, reads last quarter's answers and the policy doc, and fills the form in one pass. The nine questions it could not evidence come back flagged, each with the reason, for your AE to answer.

Making your data answerable and finishing the work inside it are two different jobs.

Microsoft documents connectors as the way Copilot reaches data beyond Microsoft 365.

Synced connectors index that data into Microsoft Graph for Copilot and search. Federated connectors fetch data in real time, without indexing content into Microsoft 365.

That is a retrieval arrangement. The data becomes something you can ask about.

Strawberry does the other half of the job. Rows arrive in your sheet with the source linked beside each value. The ticket is updated against the real ticket. The form in the prospect's portal is filled in and left for a human to check. That is what 2,000+ integrations is for, so the result lands where the work actually lives rather than in a reply you have to re-type.

Adding the tool you use should not require an admin, a license change or a rollout.

Microsoft documents three connector arrangements.

Tenant connectors are configured by an admin. Self-serve connectors are enabled by an admin, and users authenticate to them. Custom connectors are listed for tenant configuration. Microsoft Copilot itself requires an add-on license, and users must have a Microsoft 365 license assigned to them.

In Strawberry you add what you need on the afternoon you need it. You connect the app yourself and see the exact permissions before you approve them.

Apps with several services ask what your AI companion may use, so you can let it read your mail without letting it send. You can keep more than one account for the same tool and set which is the default. If the tool your team runs on is not there yet, you can add it yourself with a URL. Nobody has to approve a project for your AI companion to learn a new tool.

Nothing sends until you have seen the message, the recipient and the attachment.

Reading a policy doc, filling in a form, submitting that form and updating a live record are four different decisions. Strawberry asks you about each one separately.

Every connected action is set to ask, always allow, or off. Drafting an answer into the form can run unattended. Submitting it to the prospect does not. Payments stay off unless you turn them on yourself.

What you approve is the finished item, and you can change it there. The destination, the recipient, the wording and anything attached are all open to edit. On a reply you also see the address it resolved, and no other thread recipient is added for you.

Microsoft documents its own controls in Copilot Studio, which includes human-in-the-loop steps in workflows. It also documents event triggers, which let an agent act autonomously in response to a defined event.

Work that repeats should be a schedule you can audit, not a thing you have to build first.

Microsoft documents Copilot Studio as a low-code studio for building and managing agents and workflows. You create an agent by describing it in plain language and test it before you publish, and workflows are built in a drag-and-drop designer.

In Strawberry the thing that repeats is the AI companion you already worked with this morning. There is no second product to open. You say keep doing that. It keeps the sources you trust, the way you judge what counts, and the point where you want to look before it carries on.

Then give it a schedule, or a trigger such as a new email arriving. Every run records what it did and what it cost, so a schedule is something you can audit rather than something you hope is still working.

Then hand it over. An AI companion arrives at your teammate carrying its memory, its skills and its connected accounts. They get a colleague who already knows how your questionnaires get answered. The team workspace holds the shared files, and the team credit pool means nobody has to ask for a license first.

Frequently asked questions