MMOOZZGG.com

Why AI does not work

Custom AI development vs off the shelf: when ready-made is enough

5 min read2 sources

In custom AI development vs off the shelf, the contractor often decides for you, and usually in favor of development. Large companies in Russia decide differently: in a 2025 industry study, 78% of them use ready-made AI solutions from outside partners. Below: when a ready-made product is enough, when you cannot do without development, and how we choose.

On this page · 5 sections

Why contractors offer development

Contractors sell what they know how to build. A development team earns on hours and stages, and a customer’s subscription to a ready service earns it nothing.

Custom work looks solid: your own solution, your own code, your own team. For an owner it sounds like independence, but in practice it often means months of work and a budget that grows with each stage.

The third reason: a contractor does not have to know the market. To propose a ready product, you must follow dozens of services, their prices and limits. It is easier to offer what the team has built before.

A warning sign: the proposal contains no ready-made option to compare, and the answer to “why not use a service?” is a general reply about flexibility.

What the numbers say

The study surveyed 150 technology directors of large companies in 16 industries. 78% use ready-made AI solutions from outside partners, and 71% fine-tune ready models instead of a full training cycle.

Even companies with their own IT departments and large budgets do not build AI from scratch. They take a ready product and adapt it to their processes.

In its 2025 report, MIT NANDA compared how projects reach real use. With an outside partner this happened in about 67% of cases, and with in-house development in 33%. These are survey data, not an exact measurement, but the gap is twofold.

The lesson for an owner: a partner’s experience and ready solutions raise the chance that AI reaches a working process. Custom AI development makes sense where nothing ready exists, not where nobody looked.

Where a ready-made solution is enough

Standard processes that work the same way in thousands of companies: taking requests, booking clients, answering frequent questions, transcribing calls, screening résumés, recognizing documents.

Signs that a ready product is enough: the process is described and the information sits in standard systems such as a CRM or ERP. The market also has services that connect to them.

A ready solution launches faster, costs a subscription instead of a development budget, and has already been tested on other clients. Our overview of AI prices shows what such solutions cost.

Ready-made does not mean “buy and forget”. The solution is set up for your rules, connected to your systems and checked against a metric, like any project.

Before you choose, ask the vendor of a ready solution for a pilot on your own requests. You will see whether the service handles your cases before you pay for a year’s subscription.

Where you cannot do without development

1. The process is yours alone and gives the company an edge: its own calculation method, pricing or production logic. No ready solution exists, because nobody else has such a process.

2. The information must not leave the perimeter: customers’ personal data, trade secrets, the requirements of 152-FZ, Russia’s personal data law. Then you need a model on your servers or in a Russian cloud, connected to your systems.

3. The company’s systems are non-standard: its own accounting software, an old database, production equipment. A ready service will not connect to them without extra work.

4. A ready solution exists, but at your volume it costs more than your own development with support. A calculation proves this, not the contractor’s opinion.

Even in these cases, development is built from ready parts: the model is fine-tuned, not trained from scratch. 71% of the large companies in the study do this.

How we choose

We count money and do not sell development: first we see what will bring an effect, then we decide how to do it.

For each process we check the market: ready services, AI agents, local models. Next to them we put development and compare the options by price, term, effect in money and risks to company information.

The calculation shows where a ready solution covers the process fully, where it needs tuning, and where you cannot do without your own development. You make the decision, by the numbers, not by what the contractor’s team knows how to do.

If a contractor has already sent you a proposal, you can test it with the same questions: what the market offers, what it costs, which metric is written down and who owns the code and data. The steps are on the page about the check of an AI solution.

When a process needs an AI agent, we start with a pilot: one agent on one process, from $1,499, 14 days, with the metric in the contract. If there are several processes, we build an AI department from $4,999, 4–6 weeks: 3–5 AI agents connected to the CRM, ERP and phone system.

Sources and fact-checking

Checked by the editors on 13 October 2026.

  1. Yandex and Yakov and Partners: study of AI in Russia, 8 December 2025 (yandex.ru)
  2. MIT NANDA: The GenAI Divide, State of AI in Business 2025 (mlq.ai)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min
Contact us