How to Choose an Automation Integrator: 7 Criteria for Evaluating Robotics Partners

In a robotic automation investment the most expensive mistake is not picking the wrong camera or the wrong robot brand — it is picking the wrong company to run the project. Hardware can be swapped; six lost months cannot. Automation integrators tend to submit quotes that look remarkably alike, and the difference shows up not in the quote but at commissioning. This article covers the criteria worth applying when you evaluate automation companies, and the warning signs to catch before signing.

First, Decide Which Kind of Automation Company You Need

The phrase “automation company” covers very different business models, and going to the wrong type is the first step of most failed projects:

• Electrical and control automation: PLC programming, panel building, SCADA. Controls an existing machine; perception and decision-making are out of scope.
• Special machine builders: mechanically led design and manufacture. Strong for a repetitive, well-defined process; strained by variable part geometry.
• Robotics integrators: adapt a robot arm to a process — cell design, gripper, safety, robot programming.
• Machine vision and AI specialists: the perception and decision layer — defect detection, pose estimation, classification.

Applications such as bin picking, visual quality control or autonomous harvesting sit at the intersection of the last two. Taking a “recognise a variable part and pick it” problem to a team that only programs robots means leaving the riskiest part of the project to be discovered on site.

1. Do They Have a Reference Close to Your Application?

What matters is not the length of the reference list but its similarity. The right question is not “how many projects have you done” but “which project have you done that resembles our part type, our cycle time and our number of variants”. If possible, see a running cell in person and talk to that plant’s maintenance team — not the sales team.

2. Do They Offer a Proof of Concept?

For any project involving perception, a proof of concept is the only real risk-reduction method: a pre-study run with your actual parts, under realistic lighting, against a measurable success criterion. “This will work for us” is not a commitment. A good integrator defines the success metric with you in writing before the PoC: detection rate, false reject rate, cycle time.

Managing the transition from PoC to series production is a separate competence; we described why that transition breaks in so many projects in industrial AI process management.

3. Who Owns the Software, and Who Maintains It?

This is the clause most often skipped in contracts and most expensive later. Ask: who owns the application software? Are source code or project files handed over? Will you have to call the integrator every time a new part type is added, or can your own team be trained to do it? Who retrains the AI model?

Your team being unable to add a new variant is the single most common reason a system falls out of use within two years.

4. Brand Independence and the Reasoning Behind the Hardware

A company tied to one brand designs the solution around what that brand can do. There is a simple way to test this: ask why the proposed camera or robot was selected. A good answer is grounded in the application — part size, reflectivity, depth accuracy, cycle time. A weak answer is “this is what we work with”. Our 3D camera technology comparison gives you the technical ground to ask that question properly.

5. Post-Commissioning Service and Spare Parts

When the line stops, within how many hours does someone respond? Is remote access available, and how is it secured? What is the lead time for spare parts on critical components? How large is the local service team, and do those people know your cell? Do not proceed with a company that will not put service scope in writing at quotation stage.

6. Documentation and Operator Training

The handover package should include electrical schematics, mechanical drawings, an I/O list, an operating manual and a troubleshooting guide. Operator training should be a hands-on session with real parts, not a one-off presentation, and the maintenance team needs its own technical session. In projects where documentation is promised “later”, it usually never arrives.

7. Scope Clarity and Total Cost

To make quotes comparable, ask every bidder to itemise the same lines: robot and camera hardware, gripper and fixturing, safety equipment, electrical and mechanical installation, software, PoC, commissioning, training, documentation, warranty period and annual maintenance fee. The lowest quote is usually the narrowest in scope — the difference returns mid-project as a change request.

We broke down how the real cost of an investment forms beyond hardware price in integration cost in robotic automation, and how payback is calculated in our investment and ROI guide.

Red Flags Before You Sign

• Committing to an exact cycle time and accuracy without ever seeing your parts
• Dismissing a PoC as unnecessary or “a waste of time”
• Avoiding a written definition of the success criterion
• A quote that lists hardware only, with no statement of work
• Being unable to name a comparable reference, or to let you speak with it
• Never asking about lighting, mechanical fixturing or part cleanliness

That last point is the most telling: the success of a machine vision project is largely decided by lighting and mechanical repeatability. A team that does not raise those subjects in the first meeting will be meeting them for the first time on your shop floor.

How MIS Otomasyon Works

MIS Otomasyon specialises in the machine vision and AI side, and selects robots and cameras according to what the application requires — we are not tied to a single brand. As a rule we start projects with a PoC run on your real parts against a written success criterion. Explore our solution families, or request a demo to have your application assessed for feasibility.