Skip to content
Neural Soft
Contact us Español (View this page in Spanish)

Strategy

How to choose an AI assistant for your company: ChatGPT, Copilot, Gemini or Claude

How to choose between ChatGPT, Microsoft Copilot, Gemini and Claude for your company: criteria, data requirements and a two-week trial with real tasks.

A dirt path splitting in two inside a forest.

Neural Soft 4 min read

Key points

  • Start from where the team already works; Microsoft 365 and Google Workspace have their own built-in assistant.
  • Always with a business plan, a data processing agreement and no training on your data.
  • Compare using your own tasks for two weeks, not demos or leaderboards.
  • Differences between models change every few months; training and a usage policy work for all of them.

“Which do we use, ChatGPT or Copilot?”, “What about Claude?”, “Doesn’t Gemini already come with Google?”. It is one of the questions we hear most. And while the decision is pending, the team usually uses personal accounts, with company data and no control at all.

The good news is that the big tools are all very capable. The choice depends less on which one is “the best” and more on where your team works, what data it will handle and which tasks it wants to solve.

Visual guide · Choosing an AI assistant

Which AI
for your
company?

There is no single answer. The best line usually starts from where your team already works.

01 · The map

Four lines, one destination

1 You work in Microsoft 365

Copilot Chat Microsoft 365 Copilot Agents with Copilot Studio

2 You work in Google Workspace

Gemini in Workspace Gemini in Gmail, Docs and Meet Agents and automations

3 You want a general assistant

ChatGPT or Claude, business plan Projects with your documents Connectors to your tools

4 You need something your own

Your chosen model, via API With your data and systems Custom assistant
They all end at: usage policy, training and measuring use

02 · The ticket

What is non-negotiable, whichever you choose

  • Business plan

    No personal accounts with work data.

  • No training on your data

    The default on business plans; check the contract.

  • Processor agreement

    A data processing agreement (GDPR).

  • Administration

    Joiners, leavers and company sign-in.

  • Where data is processed

    Ideally in the EU; ask what each one offers.

03 · Before deciding

Two weeks with your own tasks

  1. Day 1 10 real team tasks and 2-3 candidate tools
  2. Weeks 1-2 Same tasks, same people, each tool
  3. Day 14 Score with your own criteria and decide

04 · The takeaway

The best tool
is the one your team uses well

The differences between the big assistants change every few months. Training, a usage policy and tidy data work for all of them.

Neural Soft

neuralsoft.es

Download the infographic (PNG)

First: where does your team work?

  • Microsoft 365 (Outlook, Teams, SharePoint): the natural route is Copilot. Copilot Chat comes with work accounts, and the Microsoft 365 Copilot licence adds emails, meetings and documents. We explain it in detail in Microsoft 365 Copilot in your company.
  • Google Workspace (Gmail, Drive, Meet): Gemini is built into the Workspace business plans.
  • A general assistant, separate from your office suite: ChatGPT or Claude on their business plans, with projects for working with your documents and connectors to other tools.
  • Something very specific (with your data, in your applications): a custom assistant with the best-fitting model, via API.

What is non-negotiable

Whichever you choose, check these points before giving the team access:

  • Business plan

    Accounts managed by the company, not personal ones. With joiners, leavers and company sign-in.

  • No training on your data

    The big vendors’ business plans do not use your conversations for training by default. Confirm it in the terms.

  • Contract and location

    A data processing agreement (GDPR) and, where possible, processing or storage in the EU.

The criteria for comparing

With the basics covered, the comparison has to be done with your tasks. These are the criteria that usually matter most:

  • Quality on your tasks

    Writing in your tone, summarising your documents, analysing your data. Not in a generic demo.

  • Integration

    Whether it works inside the tools you already use or forces copy and paste.

  • Data and security

    What information it can reach, with what permissions and what controls the administrator has.

  • Total cost

    Per-user licences, who really needs them and the time for training and roll-out.

To structure the decision, a weighted scorecard stops the tool that impressed most in the first test from winning by default:

Criterion Weight Tool A Tool B Tool C
Quality on our tasks ×3 4 / 5 5 / 5 4 / 5
Integration with what we use ×3 5 / 5 3 / 5 2 / 5
Data and security ×2 4 / 5 4 / 5 3 / 5
Ease of use for the team ×2 4 / 5 4 / 5 3 / 5
Total cost ×1 3 / 5 4 / 5 5 / 5
Weighted total 4.2 4.0 3.2
An example with fictitious data: each company sets the criteria, weights and scores after the trial.

How to run the trial in two weeks

  1. Choose 10 real tasks

    The ones the team does every week: a difficult email, a report, summarising a contract, analysing a spreadsheet.

    List of test tasks

  2. Choose 2 or 3 candidates

    Usually the one built into your office suite and one or two general assistants, all on a business plan or a business trial.

  3. Same tasks, same people

    Each person tries the same tasks in each tool and notes the result and the time taken.

  4. Score with your criteria

    Using the weights agreed before starting, not afterwards.

    Scorecard

  5. Decide and train

    One approved tool (or two, by role), a usage policy and short training for the whole team.

One tool for everyone or several?

One for everyone

  • Simpler to manage and to train.
  • One policy and one contract.
  • Enough for most SMEs.

Several by role

  • The built-in one for everyone and another for specific roles.
  • Useful when tasks differ a lot (development, analysis, writing).
  • Needs more discipline: all approved and in the usage policy.

Next step

In an AI consulting engagement we prepare the trial with your tasks, review contracts and data, and support you through the roll-out. For the team, the practical AI at work course works with any tool, and there are specific courses on Microsoft 365 Copilot and Google Workspace with Gemini.

Frequently asked questions

Can we use more than one tool?
Yes, and it is common: for example, the assistant built into Microsoft 365 for everyone and another general assistant for people who write or analyse a lot. What matters is that they are all approved, on a business plan and covered by the usage policy.
Are free versions fine for work?
Not for company data. Personal versions may use conversations to improve models and give the company no control over accounts. For public information the risk is lower, but the whole team should use the approved tools.
Which is the best model?
It depends on the task and changes often: every few months a new version overtakes the others at something. That is why we recommend deciding with your real tasks and reviewing the choice every year.

Related services

  • AI consulting for SMEs

    We help small and medium-sized businesses find where AI can add value, which cases are worth testing and how to start without investing in solutions they do not need.

  • Chatbots and internal assistants

    We build chatbots and internal assistants connected to your company's information to answer questions, reduce repetitive queries and help the team find what it needs faster.

Related training

More articles

Want to apply this in your company?

Tell us about your case and we will tell you, with no commitment, whether AI makes sense and where to start.

Request an AI assessment