Cloud, in the EU
The most capable commercial models, served from the EU, for the tasks that need them.
- Claude
- GPT
- Gemini
- Mistral
The everything app for business
Teler, where people and AI experts work together
One place where the team talks to each other, to customers and to AI experts, and where agents do the work on top of the company’s systems. They read email and WhatsApp, enter orders, draft proposals and ask for your go-ahead before they touch anything.
Sample company · nothing to install · just your email
- 01One inboxEverything that needs your attention, whether it comes from people, customers or agents.
- 02A mixed teamPeople and AI experts with roles, tools and limits, on the same org chart.
- 03Screens on demandViews, reports and documents generated when you ask for them.
The everything app for business
Your ERP and your email. Everything else, in Teler.
Every new tool is another subscription, another screen and another thing to keep up to date. In Teler, that work is done by AI experts on top of the systems that stay.
What you can stop paying for
- CRMHubSpot · Salesforce · Pipedrive
- Team chatSlack · Teams
- Tasks & projectsAsana · Trello · Monday
- Reports & dashboardsPower BI · Looker Studio
- Proposals & documentsPandaDoc · plantilles de Word
- HelpdeskZendesk · Freshdesk · Intercom
- AutomationsZapier · Make · n8n
- Wiki & knowledgeNotion · Confluence
- Each tool’s AI assistantCopilot · Gemini
What stays
- Your ERP, the system of record
- Email: Outlook or Gmail
- WhatsApp, if you use it
| Tool | Examples | In Teler |
|---|---|---|
| CRM | HubSpot · Salesforce · Pipedrive | A customer record that updates with every conversation |
| Team chat | Slack · Teams | Channels with people and experts |
| Tasks & projects | Asana · Trello · Monday | Assignments in the inbox, with an owner and a trail |
| Reports & dashboards | Power BI · Looker Studio | Views generated on request, with live data |
| Proposals & documents | PandaDoc · plantilles de Word | Proposals and reports generated from context |
| Helpdesk | Zendesk · Freshdesk · Intercom | A portal where an expert answers with real data |
| Automations | Zapier · Make · n8n | Experts that work across systems and handle the exceptions |
| Wiki & knowledge | Notion · Confluence | One company context, read by people and agents |
| Each tool’s AI assistant | Copilot · Gemini | Model-neutral AI experts across every system |
Each company decides what to drop and when. Teler also connects to the tools you want to keep.
Where Teler fits
Teler lives between people and systems
At the top, the team. In the middle, Teler: the workspace, the AI experts and the company context. At the bottom, the ERP and the usual systems, which stay. The numbers show how each part interacts.
How people and experts interact
| # | Who with whom | Where it happens | Example | Who decides |
|---|---|---|---|---|
| 1 | Person with person | Channels and messages; @ brings in whoever is needed | Jordi and Laia review the Grues Penedès offer | The people |
| 2 | Person to expert | “Assign a task”, in the inbox or on mobile | “Enter Ferreteria Puig’s order”, to Arnau | Whoever assigns it |
| 3 | Expert to person | Work prepared in “For you”, with the reason | Purchase order PO-7790 to Festo, waiting for Núria’s confirmation | The person, always |
| 4 | Expert with expert | The orchestrator, within the same context | Nil brings the history and Clara drafts the offer | Each expert’s limits |
| 5 | Customer with the company | Email, WhatsApp or portal, in each customer’s thread | Ferreteria Puig orders on WhatsApp and gets a confirmation | The expert, within limits; otherwise a person |
| 6 | Supplier with the company | Email and delivery notes, tied to the purchase order | Festo confirms the lead time and the delivery note is matched to the order | The person, if it doesn’t match |
| 7 | Agent with agent | An open protocol between companies (A2A) | A customer’s purchasing agent asks Clara for an offer | A person approves it |
| 8 | Experts with the ERP | Connectors, always through the agreed path | Arnau writes the order to Odoo, with audit entry #4471 | The expert’s owner |
| – | Person with the ERP | Directly, if they want (dashed line) | Accounts closes the month in the ERP, as always | The person |
Everyone talks to whoever they need. Experts do the work across systems, and every decision is written down.
AI experts
Digital workers who join your team
They aren’t tools someone opens: they’re colleagues with a trade. Each expert has its tools, its limits and a person responsible for it. You assign it work like any colleague, it runs its scheduled jobs and it leaves a trail of everything it does.
- Role and character
- Orders expert. Meticulous: never touches a price that doesn’t match without asking.
- Reports to
- Jordi Mas, who sets its limits and reviews its work.
- Tools
- Email, WhatsApp and Odoo. Only the ones it needs.
- Limits
- Writes to Odoo with approval. Limits widen as trust grows.
- Scheduled jobs
- Captures orders every 15 minutes, 7 am to 8 pm, and writes the daily summary at 6 pm.
- On demand
- “Enter Ferreteria Puig’s order.” “Which orders are blocked?”
- Model
- Swappable: the best or the cheapest for each task, in the cloud or on-prem.
Create an expert
Creating an expert is like writing a profile for a new colleague
Describe the job in your own words, give it the tools it needs, set its limits and choose who it reports to. No coding.
02Tools
03Limits
05Model
ArnauOrders expertDraft
- Reports to
- Jordi Mas · Admin
- Tools
- —
- Limits
- Acts with approval
- Model
- Claude · Cloud, in the EU
- Scheduled jobs
- Every 15 minutes, 7 am to 8 pm
- Typical request
- “Enter Ferreteria Puig’s order.”
…
Model-neutral
Any model. On your own server, too.
Each expert uses the model that suits each task. Every call goes through our own gateway: switching models is a setting, not a project.
On-prem
Open models on your own server or on EU GPUs, when data must not leave the company.
- Llama
- Qwen
- Mistral
- Gemma
- DeepSeek
One gateway for everything
- Routes and fallbacks for each task
- Limits and cost per expert
- A log of every call
- Your data never trains any model
Your ERP
We connect to whatever ERP you run
The ERP is the system of record, and it stays. Teler reads and writes to it through the path we agree, and every write is audited.
- SAP S/4HANA
- SAP Business One
- Microsoft Dynamics 365 Business Central
- Microsoft Dynamics NAV
- Sage 200
- Sage 50
- Sage X3
- Odoo
- Holded
- a3ERP
- a3innuva
- Ekon
- Oracle NetSuite
- Infor
- Cegid
- Distrito K
- Ahora
- Factusol
And any other ERP with an API, a database or exports.
01
API and MCP
For cloud ERPs and those with an open API.
02
Database
Direct reads and controlled writes when there’s no API.
03
Local connector
For the ERP in your office, over an encrypted tunnel.
Integrations
Connect the tools you already use
Pick one or two apps and see what an expert can do with them. These are examples: in every company, workflows follow its own rules and limits.
Pick up to two apps to see what can be done with them.
24 example workflows
- WhatsApp BusinessOdoo
WhatsApp orders, straight into the ERP
- A customer orders on WhatsApp
- Checks price, stock and risk in Odoo
- Enters it and confirms to the customer
- OutlookSAP Business One
PDF orders that arrive by email
- Reads the attached PDF
- Creates the order in SAP Business One
- Asks you if a price doesn’t match
- GmailOdooExcel
Quotes in minutes, with real data
- A quote request comes in
- Drafts the proposal with price and stock
- You review it and it goes out
- DocuSignBusiness Central
Proposal signed, order created
- The customer signs in DocuSign
- Creates the order in Business Central
- Tells the warehouse in the channel
- Bank (PSD2)Sage 200
Bank reconciliation every morning
- Reads last night’s transactions
- Matches them to Sage invoices
- Leaves what doesn’t match in your inbox
- HoldedGmailWhatsApp Business
Chasing overdue invoices
- Spots overdue invoices in Holded
- Drafts each reminder in the right tone
- If they don’t pay, suggests a call
- StripeHolded
Stripe payments, reconciled
- Every Stripe payment
- Marks the Holded invoice as paid
- Weekly summary in the finance channel
- OdooOutlook
Purchases before stock runs out
- An item drops below minimum in Odoo
- Proposes the purchase with usage and lead times
- You confirm and it goes to the supplier
- Google SheetsSAP Business One
From sales forecast to purchase plan
- Reads the forecast in Google Sheets
- Cross-checks SAP stock
- Suggests what to buy and when
- ShopifyBusiness Central
Online shop orders into the ERP
- An order lands in Shopify
- It’s created in Business Central with customer and shipping
- The customer gets a confirmation
- WooCommerceSEURWhatsApp Business
“Where’s my order?”, answered instantly
- The customer asks on WhatsApp
- Checks status in WooCommerce and SEUR
- Replies with the tracking number
- DHLOutlookOdoo
Delivery issues without chasing anyone
- DHL reports a delay
- Tells the customer and proposes a new date
- Updates the order in Odoo
- Outlooka3ERP
Supplier invoices, matched and booked
- The invoice arrives by email
- Matches it to the order and delivery note
- Books it in a3ERP or flags the difference
- HubSpotGmail
Deal ready, proposal drafted
- A HubSpot deal moves to “proposal”
- Drafts the proposal from the history
- Leaves it ready for you to send
- PipedriveGoogle Calendar
Sales follow-up that never slips
- A proposal sent a week ago
- Checks whether the customer opened it
- Books you a call in the calendar
- LinkedInWebsite & formsSalesforce
A brief before every visit
- You have a visit tomorrow
- Gathers who’s who, what they make and what worries them
- Leaves it in the Salesforce record
- Website & formsZoho CRM
Website leads, qualified
- Someone fills in the website form
- Qualifies them with what it knows
- Creates them in Zoho and proposes a meeting
- MailchimpHubSpot
From campaign to call list
- A Mailchimp campaign ends
- Sees who opened and clicked
- Builds the call list in HubSpot
- Google AdsGoogle Sheets
Ad report every Monday
- Reads Google Ads spend and leads
- Cross-checks against real sales
- Tells you what works and what doesn’t
- ZendeskOdoo
Complaints solved with data
- A complaint lands in Zendesk
- Checks the order and delivery note in Odoo
- Proposes the credit note for approval
- Microsoft TeamsSharePoint
Meetings with minutes and tasks
- A Teams meeting ends
- Writes the minutes and assigns tasks
- Saves it to SharePoint
- GmailGoogle CalendarGoogle Drive
An email assistant for the whole team
- Sorts email by priority
- Drafts replies with the customer’s context
- Proposes meetings and attaches what’s needed from Drive
- ExcelBusiness Central
The management dashboard, no exports
- Every Monday at 8
- Sales, margins and blocked orders from Business Central
- A generated view, and the Excel if you want it
- AsanaSlack
Hours and projects, up to date
- Reads Asana tasks and conversations
- Flags projects going off track
- Drafts the month’s hours invoice
Teler connects through APIs, MCP or a connector of its own. If your tool isn’t listed, we connect it during the project.
Expert catalogue
Experts for every sector and every team
A catalogue that grows with every project. The same experts serve many companies: what changes is each one’s systems, rules and limits.
22 experts
- Finance
Bank reconciliation
Matches every transaction to invoices and receipts, and leaves you only what doesn’t match.
“Reconcile yesterday’s transactions and tell me what’s missing.”
Banca · ERP
- Finance
Collections
Watches due dates, sends reminders in each customer’s tone and tells you when to call.
“Who owes us for more than 30 days?”
ERP · correu · WhatsApp
- Finance
Supplier invoices
Reads invoices, matches them to the order and delivery note, and books them.
“Which invoices this month don’t match?”
Correu · ERP
- Operations
Order capture
Reads orders from email, WhatsApp and portal, checks price and stock, and enters them in the ERP.
“Enter Ferreteria Puig’s order.”
Correu · WhatsApp · ERP
- Operations
Stock & purchasing
Watches minimums, usage and lead times, and prepares purchases before stock runs out.
“What will run out this month?”
ERP · correu
- Operations
Logistics & deliveries
Tracks shipments, flags delays and reschedules deliveries with the customer.
“Which of today’s deliveries are late?”
ERP · transportistes · WhatsApp
- Operations
Production planning
Sequences work orders by date, materials and capacity, and flags bottlenecks.
“Can we ship 500 more units on the 20th?”
ERP · fulls de producció
- Operations
Quality & non-conformities
Logs issues, traces the cause through batches and suppliers, and drafts the report.
“Where did Tuesday’s faulty batch come from?”
Documents · ERP
- Service
Technical support
Answers questions from the datasheets, opens tickets and schedules visits.
“Which solenoid valve replaces the EV-52G14?”
Correu · ERP · fitxes tècniques
- Sales
Proposals & quotes
Drafts product or service proposals with real prices, stock and terms, and tells you what to check.
“Draft the quote for Hidràuliques Montseny.”
ERP · correu · documents
- Sales
Sales follow-up
Follows up every proposal and tells you when to call.
“Which proposals are going cold?”
CRM · correu · agenda
- Sales
Public tenders
Finds tenders that fit, summarises requirements and prepares the paperwork.
“Any construction tenders in Barcelona this month?”
Web · documents
- Operations
Progress billing
Prepares monthly progress certificates from measurements and flags overruns.
“How is the Sabadell site budget doing?”
Documents · ERP · correu
- Management
Market research
Studies competitors, prices and prospects, and sums it up with sources.
“Who sells ISO cylinders in Catalonia, and at what price?”
Web · CRM
- Management
Strategy & management
Dashboards, scenarios and board decks, whenever you ask.
“What happens to margin if we raise prices 3%?”
ERP · tot el context
- Marketing
Marketing & campaigns
Plans campaigns, measures them against real sales and suggests what to adjust.
“Which campaign brought us the most orders?”
Web · correu · anuncis
- Marketing
Content & GEO
Keeps your website readable for people and buying agents, so AI engines cite you.
“Do we show up when someone asks ChatGPT?”
Web · cercadors d’IA
- Service
Email assistant
Sorts email, drafts replies with context and proposes meetings.
“What do I absolutely have to answer today?”
Correu · agenda · documents
- Service
Customer service
Answers order status, lead times and questions at any hour, and escalates what’s needed.
“Which of today’s queries are unanswered?”
WhatsApp · portal · ERP
- Sales
Bookings & events
Drafts event quotes with menus, allergens and availability, and follows them up.
“Quote a dinner for 80 people on the 14th.”
Correu · agenda · ERP
- Operations
Contracts & rights
Reads contracts, extracts terms, territories and conditions, and warns you before anything expires.
“Which rights expire this year?”
Contractes · documents
- Finance
Project hours & billing
Gathers project hours, flags overruns and drafts the invoice.
“Which projects are over budget?”
Projectes · ERP
How it works
Five layers and one context
01
One context
People and experts work with the same information. What one expert learns, the whole team benefits from.
02
The ERP rules its own domain
Quantities, invoices and accounting stay in the ERP. Teler doesn’t replace or migrate it.
03
Interchangeable surfaces
The inbox, the portal or the website can be redesigned without touching data or experts.
The journey of a task, step by step
An order arriving on WhatsApp at Torrent Components. People only see what their decision needs.
Dashed line: if the order is within the limits Jordi has set, Arnau enters it himself and explains it in the channel. Solid line: if a price doesn’t match or a limit is exceeded, the decision lands in the inbox with all the work done.
People decide; experts execute
Each type of action has an autonomy level, set by the person responsible for the expert. The level rises as the expert earns trust.
- 01
Informs
The expert flags it and the person acts.
ExampleMarc Serra has opened the proposal again.
- 02
Proposes
It prepares the work and the person chooses.
ExampleTwo options: 60 days, or 90 with a 2% surcharge.
- 03
Acts with approval
It gets everything ready and the person confirms in one click.
ExamplePurchase order PO-7790 to Festo, waiting for Núria’s confirmation.
- 04
Acts and reports
It does it alone, within limits, and explains what it did.
ExampleSummary of the day’s orders in the #orders channel.
Built-in controls
- Permissions per expert and per tool
- Spending limits
- Approvals in the inbox
- Audit of every action
- Cost per expert
- Instant pause
Data, in a language built for agents
Today, information lives on screens, in spreadsheets and in people’s heads. In Teler every fact, document and decision is stored in a format an agent can read and a person can understand.
Facts, in JSON
{
"client": "Grues Penedès",
"termini_dies": 90,
"recarrec": 0.02,
"impagats_4_anys": 0,
"aprovat_per": "Jordi Mas"
}Knowledge, in Markdown
## Payment terms Customers with no late payments in 3 years: up to 90 days, with a 2% surcharge.
Decisions, with a trace
Who, what, when and why: approved by Jordi Mas at 09:58, with the payment history as the reason. The data belongs to the company and exports in open formats.
01
All in the EU
Data and models in the EU. GDPR by design, on the same stack we already use for client projects.
02
Each company, isolated
Row-level isolation in the database and separate indexes per company. Your data never trains any model.
03
No vendor lock-in
Your data is yours and exports in open formats. The model is whichever fits best.
The product
The day starts in the inbox
- 01For youWhat’s waiting for your decision, in order of priority.
- 02Team channelsSales, Operations and Admin, with people and experts in the same thread.
- 03AI expertsHired like any team member, and one message away.
- 04Customers and suppliersEmail and WhatsApp land here, linked to each customer’s record.
- 05The conversationJordi asks; Nil answers with the payment history and brings in Clara.
- 06DoneClosing a task is one click, and it’s recorded.
Screens are generated when you need them
When Núria wants to check stock, she asks. Teler opens a view built for that question, with live data from the ERP. When she’s done, it closes. Every write to the ERP leaves a trail: what changed, who approved it and why.
- 01
Ask
“How are we doing on cylinder stock?”
- 02
Query
The expert reads Odoo through the agreed path.
- 03
View
A screen is generated for that question.
- 04
Decide
Núria confirms the purchase from Festo.
- 05
Trail
It’s written to Odoo, with an audit entry.
What changes in the company
Word, Excel and PowerPoint files still exist as delivery formats. What changes is where the work happens.
| Today | With Teler | The person |
|---|---|---|
| Email and WhatsApp, each in its own app | One inbox: every message, with the customer and their context | Answers what needs judgement |
| Spreadsheets exported and cross-checked | Live data and views generated on request | Asks and decides |
| Documents and decks made by hand | Proposals, reports and decks generated from context | Reviews and sends |
| A CRM updated when someone has time | A customer record that updates with every conversation | Calls and negotiates |
| Team chat in Teams or Slack | Channels with people and experts | Coordinates |
| A portal or forms for customers | A portal where an expert answers with real data | Handles the exceptions |
| Month-end reports | Answers and dashboards on the spot | Acts |
| ERP: Sage, Odoo, Business Central… | Stays. Teler reads and writes to it through the agreed path | Goes in when they want, not to retype |
A working day
A day at Torrent Components
A pneumatic components distributor with warehouses in Granollers and Rubí: 25 people, Odoo 17 and a team of AI experts. The company is a sample; every scene comes from a real client case.
- Arnau · orders
Has read 23 emails and 9 WhatsApps from overnight. 14 orders entered in Odoo; 2 sent to the inbox because a price doesn’t match.
Odoo · inbox - Marta · collections
Has reconciled the bank. Three bounced direct debits: the reminders are ready for Jordi to send.
Bank · inbox - Nil & Clara · sales
Grues Penedès asks for 90 days. Nil brings the payment history and Clara drafts two options.
Conversation - Pol · purchasing
Cylinder CIL-63200 drops below minimum. He leaves the Festo purchase waiting for Núria’s confirmation.
Odoo · audit - Nil · heads-up
Marc Serra opens the proposal again. Nil tells Laia: call him now, it’s on his mind.
Inbox - Management
“How are we doing this month versus last year?” A view generated with sales, margins and blocked orders.
On-demand view - Arnau · summary
Daily summary in the #orders channel: 38 tasks done and both exceptions resolved.
#orders channel
Nobody opened the ERP to retype anything. The team spent the day deciding, negotiating and calling.
Live this day in the demoAlready working, in pieces
Every Teler module comes from a project a client has paid for and uses
Teler brings them together in one system.
In use
AI-powered CRM
Four companies in different sectors work with custom CRMs built by Impuls.cat.
In user testing
Contract reading
950 audiovisual rights contracts, from 2005 to 2026, turned into searchable data. Seven agents planned.
In use
Industrial content & GEO
Technical content and structured data so AI engines find and cite an industrial manufacturer.
In use
Conversational assistants
Round-the-clock support and lead qualification with the company’s knowledge, in healthcare.
In progress
Agentic websites & back office
Websites with a CMS and their own agents, and the Teler back office at a coworking client.
Proposed
Layers on top of the ERP
Orders and a B2B portal on Sage, and an invoice agent on Business Central.
The rule: no module is built to fill a catalogue. It’s built when a client pays for it and uses it.
Why now
The shift, in numbers
≈ 1,200
app switches a day per person. Almost 4 hours a week just to get back on track.
HBR, 2022
2 min
between interruptions: meetings, emails or messages.
Microsoft, Work Trend Index 2025
60%
of Spanish SMEs still process invoices by hand.
Sage and Afi, 2026
10,000+
public MCP servers, the open standard for agents to use tools.
Linux Foundation, Dec 2025
−25%
the S&P software and services index between 12 January and 23 February 2026, after the launch of Claude Cowork.
DeepLearning.AI, February 2026
20%
of Spanish companies with 10+ people use AI; 80% still don’t.
Eurostat, December 2025
The very notion of business applications may collapse in the agent era: the logic moves to the AI layer, which updates the databases directly.
By 2028, a third of user experiences will shift from native applications to agentic front ends.
Who’s in the game, and what an SME is missing
They all apply AI to the system of an era when people worked in front of screens. Teler starts from the other side: agents do the work and people are there to decide.
| Category | Examples | Where the AI works | What an SME is missing |
|---|---|---|---|
| Office assistants | Microsoft 365 Copilot, Gemini in Google Workspace | Inside Word, Excel, Outlook or Gmail | People still open each app and bridge them all. |
| Agents inside the ERP | Sage Copilot, SAP Joule, Business Central agents | Inside each vendor’s ERP | Each vendor works within its own walls; email, WhatsApp and Excel are left out. |
| Agent platforms | Slack with Agentforce, OpenAI Frontier, Microsoft Agent 365 | On top of each giant’s ecosystem | They need a technical team to deploy, and no one rolls them out alongside the SME. |
| Single-role workers | 11x, Artisan | In one isolated process, such as prospecting | One role on its own, with no shared space with the team and no company context. |
| Automations | n8n, Make, Zapier | In fixed flows between apps | Rules set in advance; when an exception comes up, it goes back to a person. |
| Teler | Impuls.cat | In an inbox where people and experts work, across every system | Vendor- and model-neutral, built for SMEs and rolled out alongside the client. |
Categories describe where the AI works in each case, not the quality of each product.
The risk
Gartner expects more than 40% of agent projects to be cancelled by the end of 2027, due to costs, unclear value or lack of control. Teler puts context, limits and a responsible person for each expert at the centre.
Roadmap
Above all, a framework for the AI era
Today Impuls.cat builds each expert. Tomorrow partners and companies themselves will build them, with the same tools we use.
01
Open to build on
An expert kit, MCP for tools and A2A to talk to other companies’ agents: open standards from the Linux Foundation.
02
Governed by default
Everything built on it inherits the core’s permissions, limits, approvals and audit.
03
Model-neutral
Each expert uses the model that suits it, always through the same gateway.
The product grows inside real projects
A build partner is a company that commissions a project from us below market price. In return, the software it produces becomes part of Teler.
The client keeps
- Its data, which never leaves its space and exports whenever it wants.
- The processes and rules that make it different. They’re never reused.
Teler keeps
- The software: experts, connectors and views, generalised for any company.
- Each expert’s pattern, so the next client doesn’t start from scratch.
From projects to a framework, in four stages
- 2026 · now
Build partner projects
AI layers on each client’s ERP. Every project leaves a reusable module.
- 2027
Teler Core
Inbox, experts, context, connectors and governance on one base. Commercial launch; build partners become the first customers.
- 2027–2028
Catalogue & partners
Experts by function and verticals by sector. “Hire an expert” from the catalogue.
- 2028 onwards
Open framework
A kit to build experts, connectors and views on Teler, and agents from different companies that talk to each other.
Indicative timeline. Each stage opens when the previous one works with real clients.
Build partners
We’re looking for companies that want to build it with us
What you bring
- A process that costs you hours today: orders, invoices, proposals or collections.
- Access to your systems: ERP, email and documents, through the path we agree.
- A responsible person who decides with us and is the first user.
What you get
- The system rolled out below market price.
- Founder terms when Teler launches.
- A direct say in the product and the roadmap.
- Your data and your processes, which are yours and are never reused.
Teler will be the place where the work gets done. We’re building it with the people who do that work every day.
Questions
What people usually ask.
What is an AI expert?
A digital worker that joins your team. It has a trade (orders, proposals, purchasing, collections…), specific tools, limits and a person responsible for it. You assign it work like a colleague, it runs scheduled jobs and it leaves a trail of everything it does.
Is it just another SaaS?
No. Teler is the work layer that replaces many of the tools you pay for separately today. Your ERP and your email or WhatsApp stay; the rest of the work is done by AI experts inside Teler, following your rules.
Do I have to change my ERP?
No. Your ERP stays: it’s the system of record. Teler reads and writes to it through the path we agree (SAP, Business Central, Sage, Odoo, Holded, a3…) and every write is audited.
Which AI models does it use? Can they run on-prem?
Whichever fits each task best: Claude, GPT, Gemini or Mistral in the EU cloud, or open models such as Llama, Qwen or Gemma on your own server. Everything goes through a gateway, so switching models is a setting.
Where is my data?
In the EU. Each company is isolated in the database and your data is never used to train any model. It’s yours and it exports in open formats.
What if an expert makes a mistake?
That’s why nothing runs without a responsible person. Every expert has limits, important actions wait for your approval and everything goes into the audit log. And you can pause it instantly.
How much does it cost?
Right now we work with build partners: we roll Teler out on one specific process, with a project fee and a monthly subscription, below market price. We’ll give you a figure once we know which process you want to solve.
Does the demo use real data?
No. Torrent Components is a sample company. The scenes come from real client cases, but the names and figures are made up for the demo.
Who is behind Teler?
Impuls.cat, an AI consultancy in Barcelona (Norrsken House) that builds CRMs, agents and websites for B2B SMEs. Teler brings together what we already do in client projects.
See how a mixed team works.
A sample company, five minutes and just your email.
Sources
Consulted in October 2026. Market figures come from the sources cited; Torrent Components and its data are an illustrative example.
- HBR (Murty, Dadlani and Das), app switching, via Canadian Underwriter · Aug 2022
- Microsoft, The 2025 Annual Work Trend Index: The Frontier Firm is born · Apr 2025
- Portal ERP, Sage adds AI agents (includes the Sage and Afi study) · Apr 2026
- CX Today, Satya Nadella on the BG2 podcast about agents and SaaS · Dec 2024
- Gartner, 40% of enterprise apps with agents by 2026 · Aug 2025
- Constellation Research, Microsoft Agent 365 and agents in Word, Excel and PowerPoint · Nov 2025
- Microsoft Learn, Business Central Sales Order Agent · GA Nov 2025
- CIO, SAP Joule Work and the Autonomous Suite agents · Oct 2026
- No Jitter, Microsoft 365 Copilot hits 20 million seats · May 2026
- Reworked, Slack as the agentic OS at Dreamforce · Oct 2025
- ContentGrip, OpenAI Frontier for enterprise agents · Feb 2026
- METR, Measuring AI ability to complete long tasks · Mar 2025
- Epoch AI, LLM inference price trends · Mar 2025
- Anthropic, MCP and the Agentic AI Foundation (Linux Foundation) · Dec 2025
- DeepLearning.AI, Claude Cowork plugins and the SaaS sell-off · Feb 2026
- Europa Press, Eurostat data on AI use in companies · Dec 2025
- Outlook Business, Gartner on cancelled agent projects · Jun 2025
- Virtualization Review, MIT NANDA’s GenAI Divide report · Aug 2025
- Google Research, Generative UI · Nov 2025
- Ashu Garg (Foundation Capital), AI’s trillion-dollar opportunity: context graphs · Dec 2025
- Latent Space, Andrej Karpathy: Software 3.0 and the autonomy slider · Jun 2025
- Foundation Capital, A System of Agents brings Service-as-Software to life · Oct 2024
- BleepingComputer, the A2A protocol moves to the Linux Foundation · Jun 2025