In 2026 Anthropic renewed its entire Claude model family (Opus 5.5, Sonnet 5.5 and Haiku 5.5, between September and October), turned Claude into an assistant that works with your files and spreadsheets until it delivers the finished document, and updated its usage rules. For a company in Mexico, the practical takeaways are: use a business plan (Team or Enterprise) instead of personal accounts, because there Anthropic does not train on your data by default, and start with document and analysis tasks before connecting it to your systems.
What are Anthropic and Claude?
Anthropic is an artificial intelligence company; Claude is its assistant and its family of AI models. You can use it in two ways: as an application (web, desktop and mobile) that your team uses directly, or through its API, which is what systems that connect to Claude under the hood use, such as a customer service agent or an automation. Mexico is on the official list of countries where Anthropic offers its products.
NIQERA works alongside Anthropic. This article summarizes what it officially announced in 2026 that is useful to a company, with its limits. If you have not yet defined how to get started with AI, begin with how to implement AI without taking risks.
What did Anthropic launch in 2026?
These are the officially dated announcements that matter most to a company (not to a programmer):
| Date | Announcement | What it means for your company |
|---|---|---|
| February 5 | Claude Opus 4.6 | Improvements for financial analysis, documents, spreadsheets and presentations; Claude inside PowerPoint in preview for Max, Team and Enterprise plans |
| February 24 | Private plugin catalogs in Cowork | The administrator decides which plugins the team uses |
| April 9 | Cowork for enterprise deployment | Control access, spending and usage across the whole organization |
| August 26 | Built-in browser in Cowork | Claude can open sites, read pages and fill out forms as part of a task |
| September 22 | Claude Opus 5.5 | The most capable model; Anthropic reports it is much less prone to hard-to-reverse actions |
| September 28 | Claude Sonnet 5.5 | Polished documents, presentations and spreadsheets, at the same price as its predecessor |
| October 7 | Claude Haiku 5.5 | The cheapest and fastest model, designed for high volume and live support |
| October 8 | Usage Policy update | New rules; they take effect on November 12, 2026 |
| October 22 | Agent Skills moves to a neutral foundation | The "skills" you write are not tied to a single vendor |
| October 23 | Chat and Cowork unified in Enterprise | A single experience; work continues in the cloud even if you close your laptop |
Which Claude model suits a company?
It depends on the task, and it is almost never the most expensive one for everything. Anthropic describes the 5.5 family this way:
- Opus 5.5: the most capable. Anthropic says it puts the most important information first and is more resistant to prompt injection than its predecessor. For complex analysis, executive reports and long tasks.
- Sonnet 5.5: the everyday one. According to Anthropic, it is strong at well-defined tasks and at creating polished documents, presentations and spreadsheets. In an internal test with quarterly materials from a public company, two experts judged its first draft ready to send; that is the manufacturer's own result.
- Haiku 5.5: the cheapest and fastest. Anthropic designed it for high-volume, cost-sensitive tasks, and says it works especially well in live customer support.
All three are available on Amazon Web Services, Google Cloud and Microsoft Azure, in addition to Anthropic's platform. If your company already has infrastructure on AWS, you can use Claude from there.
What is Cowork and what is it good for in a company?
It is Claude working directly in the folders and tools you choose, until it hands you the work for review. It no longer just chats: it does the whole task. The examples Anthropic gives are very office-oriented:
- You give it a folder with exports by region and budget files, and it returns a spreadsheet with the variances flagged.
- It reviews a folder of contracts against your criteria and delivers a memo per contract, with each difference cited to the clause.
- For a recurring report, it gathers the figures, compares periods and builds the presentation.
It works with Word, PDF, Excel, PowerPoint and images. On security, Anthropic states that deleting anything requires your approval, that it can wait for your go-ahead before any significant action and that you can stop it at any time. Computer control remains in research preview. Since October 23, 2026, in Enterprise chat and Cowork are a single experience; on the other plans it is rolling out.
What not supervising it costs you: a report that "looks good" with a wrong figure reaches the board meeting. Human review is still your job.
How much does Claude cost for a company?
There are two ways to pay: per user (plans) or by usage (API). Prices published by Anthropic in US dollars, before taxes, accessed in November 2026; they may change.
| Plan | Price | Who it is for |
|---|---|---|
| Team, standard seat | US$20 per user per month (annual) or US$25 (monthly) | Teams of 2 to 150 people |
| Team, premium seat | US$100 per user per month (annual) or US$125 (monthly) | Heavy users |
| Enterprise | US$20 per user per month (annual), minimum 20, plus usage at API rates | Companies that need auditing, role-based permissions and custom retention |
Through the API you pay per million tokens (text fragments: what you send is input and what Claude writes is output). Opus 5.5 costs US$4 input and US$20 output; Sonnet 5.5, US$2 and US$10; Haiku 5.5, US$0.10 and US$0.50 on requests of up to 100 thousand tokens.
Does Anthropic use my company's data to train Claude?
In commercial products, not by default. Its privacy center states that it does not use the inputs or outputs of its commercial products (Claude for business and the API) to train models, unless you submit feedback or reports, or authorize it. The Team plan explicitly includes "no training on your content by default."
On personal accounts (Free, Pro, Max) it is different: the user chooses whether to allow training, and if they allow it, Anthropic may retain their de-identified data for up to 5 years. That is why your team should not use personal accounts with customer information.
- API: Anthropic deletes inputs and outputs within 30 days, with exceptions.
- Conversations flagged by its safety systems: retained for up to 2 years.
- Zero retention: Opus 5.5 and Sonnet 5.5 are available with zero data retention, by agreement with Anthropic.
How this fits with Mexican law we explain in how to protect your customers' data when you use AI.
What changed in Claude's usage rules?
Anthropic published an update to its Usage Policy on October 8, 2026 that takes effect on November 12. For a company, two points: in high-risk uses (health, finance, employment, legal status or essential services) it requires oversight by a person with authority to review and change Claude's recommendations, and that the affected person knows AI was used. In addition, a section was added against deceptive campaigns and artificial activity, such as fake accounts, including for commercial purposes.
If you plan to use Claude to decide on credit, collections or hiring, this applies directly.
How does Claude connect to your ERP and your systems?
With open standards. MCP (Model Context Protocol) lets Claude use data and tools from other systems, and Agent Skills packages instructions for specific tasks. On October 22, 2026 Anthropic donated Agent Skills to the Agentic AI Foundation, under the Linux Foundation, where it had already donated MCP; according to Anthropic, more than 70 agent platforms already integrate it.
Odoo 20 can be connected through MCP, which lets you query your ERP from an external assistant; we explain it in the AI that already comes inside Odoo 20. For an agent to serve customers, we connect it to your ERP and CRM as an intelligent agent.
When is Claude not a good fit?
- When the task only moves data from one system to another: an automation without AI is cheaper.
- When nobody will review the result: the manufacturer's reports show progress, not infallibility.
- When your team would use personal accounts with customer data to save on the business plan.
- When the decision is high-risk and you do not have a person to review it.
Frequently asked questions
Is Claude available in Mexico?
Yes. Mexico appears on the list of countries where Anthropic offers its products, both the apps and the API.
What is the difference between Claude Team and Enterprise?
Team is for 2 to 150 people and includes single sign-on and central administration. Enterprise adds role-based permissions, audit logs, custom data retention and compliance API, with a minimum of 20 users.
Can I use Claude in Excel and PowerPoint?
Cowork works with Excel and PowerPoint files, and since February 2026 Claude inside PowerPoint is in preview for Max, Team and Enterprise plans.
Which model do I use for an agent that serves customers?
Anthropic recommends Haiku 5.5 for fast tasks such as live support; for complex cases, Sonnet 5.5. The right approach is to test with your real conversations.
Can Claude make mistakes?
Yes. That is why it should answer with data from your systems, a person should review what is important and its actions should be logged.
Sources: Anthropic, «Introducing Claude Opus 4.6», 5 de febrero de 2026 · Claude, «Cowork», 2026 · Anthropic, «Introducing Claude Opus 5.5», 22 de septiembre de 2026 · Anthropic, «Introducing Claude Sonnet 5.5», 28 de septiembre de 2026 · Anthropic, «Introducing Claude Haiku 5.5», 7 de octubre de 2026 · Anthropic, «2026 Usage Policy update», 8 de octubre de 2026 · Anthropic, «Donating Agent Skills to the Agentic AI Foundation», 22 de octubre de 2026 · Claude, «Claude Cowork and chat come together for Enterprise», 23 de octubre de 2026 · Claude, «Pricing», accessed November 2026 · Anthropic Privacy Center, «Is my data used for model training?», agosto de 2026 · Anthropic Privacy Center, «How long do you store my data?», julio de 2026 · Anthropic Privacy Center, «How long do you store my organization's data?», julio de 2026 · Anthropic, «Supported countries and regions»
