murmur

murmur · sector study · September 2026

What AI models answer when a company looks for someone to build its app or system

25 questions asked in 3 AI tools, 74 answers saved with a screenshot. No question includes a company name: these are the questions a manager types before hiring a software house.

This study does not pick a winner. The count is the same for everyone: the name is written in the answer or it is not. Anyone appearing zero times is not included on this page, and no company commissioned, reviewed or paid for this study.

01 · who appears

Nobody reaches a quarter of the answers

This is the study with the most companies cited that we have ever done. AI models wrote the name of 86 different companies in custom development, and the most cited, CI&T, appears in 16 of the 74 answers. The second, Stefanini, in 12. There is no owner of the answer: 53 of the 86 companies appear only once.

That also means many answers come out with no name at all. In 27 of the 74 the AI answers the question without naming a company; in 4, it names only one; and in 43, it lists two or more.

companyin how many of the 74 answersshare of all mentionsaverage place in the listin how many of the 25 questions
CI&T16 22%8.6%1.46
Stefanini12 16%6.5%1.95
DB18 11%4.3%4.14
Objective8 11%4.3%3.85
Ilegra7 9%3.8%5.65
Zup7 9%3.8%3.75
Ateliware6 8%3.2%5.05
BRQ5 7%2.7%3.23
SoftDesign5 7%2.7%4.04
UDS Tecnologia5 7%2.7%4.44
DevCore4 5%2.2%2.23
NextAge4 5%2.2%2.83
Accenture3 4%1.6%2.33
GFT3 4%1.6%4.32
Luby3 4%1.6%1.73
Avanade2 3%1.1%5.02
BH Softwares2 3%1.1%2.02
Capgemini2 3%1.1%4.02
DBServer2 3%1.1%7.52
Devio2 3%1.1%3.52
Dynamica Soft2 3%1.1%1.02
Elo Group2 3%1.1%6.02
Espresso Labs2 3%1.1%2.02
FWC Tecnologia2 3%1.1%1.51
Globant2 3%1.1%8.52
IBM Consulting2 3%1.1%3.52
Leven Tecnologia2 3%1.1%6.52
Mestres da Web2 3%1.1%2.52
Nava2 3%1.1%2.51
Nucleus2 3%1.1%1.02
TCS2 3%1.1%3.02
Thoughtworks2 3%1.1%3.52
Visie2 3%1.1%2.02
AGROETI1 1%0.5%1.01
Agentico1 1%0.5%2.01
AgiliSoft1 1%0.5%2.01
Agência Xará1 1%0.5%2.01
Avia Hub1 1%0.5%1.01
BeSoft1 1%0.5%2.01
Bugee1 1%0.5%1.01
CWI Software1 1%0.5%5.01
Compass UOL1 1%0.5%6.01
DF Informática1 1%0.5%6.01
DMTN Sistemas1 1%0.5%2.01
Doois1 1%0.5%2.01
EMS Tech1 1%0.5%2.01
Ecenture1 1%0.5%2.01
FCamara1 1%0.5%4.01
FeraWeb1 1%0.5%3.01
GSW1 1%0.5%1.01
GlassAuto1 1%0.5%3.01
GreenDuck1 1%0.5%3.01
IB Software1 1%0.5%2.01
IWA71 1%0.5%3.01
Ilios Sistemas1 1%0.5%2.01
LogAp1 1%0.5%1.01
Lumini IT1 1%0.5%3.01
MB Code1 1%0.5%1.01
MVP Plus1 1%0.5%4.01
Mind Group1 1%0.5%5.01
MoreApps1 1%0.5%3.01
Nano Incub1 1%0.5%2.01
Nerilabs1 1%0.5%3.01
Neryx1 1%0.5%1.01
NuageIT1 1%0.5%1.01
Nuvem Lab1 1%0.5%2.01
OCA Software House1 1%0.5%2.01
Opus Software1 1%0.5%9.01
Pitang1 1%0.5%7.01
Qigger1 1%0.5%1.01
Radix1 1%0.5%8.01
STR Software1 1%0.5%1.01
Sinapse Corp1 1%0.5%1.01
SoftNexus1 1%0.5%1.01
StudioDev1 1%0.5%3.01
Supero1 1%0.5%2.01
Sydy1 1%0.5%2.01
Target Work1 1%0.5%1.01
Tech Coders1 1%0.5%3.01
Tinnova1 1%0.5%1.01
Tria Software1 1%0.5%1.01
Usemobile1 1%0.5%3.01
Vetor Studio1 1%0.5%3.01
Viva Tech1 1%0.5%2.01
YAITEC1 1%0.5%4.01
Zug Tech1 1%0.5%4.01
How to read it. The second column is how many answers the name appeared in. The third is the share the company takes of everything that was cited: the 86 companies add up to 185 mentions and this column divides that pie. The fourth is where in the list the name usually appears: 1.0 would be always the first name.
02 · the question category

The generic question belongs to the giants; in the specific one, nobody dominates

The 25 questions were written in four categories: general recommendation, by client segment, by project need and informational. This is where the most useful data on this page is.

In the generic question, from someone who just wants the best software houses in the country, CI&T appears in 100% of the 15 answers and Stefanini in 80%. When the question narrows, that picture disappears. By segment (fintech, healthcare, agribusiness, startup, logistics), the most cited company appears in only 14% of the answers, and CI&T and Stefanini do not appear a single time. By need (legacy system, outsourced squad, AI agents, cloud), the top name sits at 12%.

question category1st2nd3rd4th
general recommendation 15 answersCI&T 100%Stefanini 80%DB1 53%Objective 53%
by segment 21 answersNextAge 14%BH Softwares 10%FWC Tecnologia 10%AGROETI 5%
by need 24 answersAccenture 12%DevCore 12%Avanade 8%Capgemini 8%
informational 14 answers

And the names change along with it. In the by-segment category, the top spots go to companies the generic question does not cite, such as NextAge and BH Softwares. For a mid-sized software house, that is the whole point: it will not compete for the generic question with a multinational, but the question for its segment is open. In the 14 informational answers, from people asking how much an app costs or how long it takes, AI does not recommend any company.

03 · one tool is not another

Each tool gets it wrong in a different way

ChatGPT is the one that writes the most names: 77 mentions, against 50 in Claude. But the volume comes with two flaws, and both are its own.

It answers by the city of whoever is asking. The machine that asked the questions is in São José do Rio Preto, and in 4 answers ChatGPT wrote sentences like "encontrei empresas de São José do Rio Preto e outras especializadas" (I found companies from São José do Rio Preto and other specialized ones), listing software houses from the city among the national ones. These companies stay out of the table, because they appeared due to the machine's address and not the question: Geocorp (2), APP Sistemas (1), Bricsoft (1), Field Control (1), RTX Tecnologia (1), Savra Tech (1), desenvolveu TECH (1).

It cites companies that do not exist under that name. 5 names written by ChatGPT do not match any software company that fits what it describes: AZO Sistemas (1), DB Tecnologia (1), Meltech (1), Novais Engenharia (1), Rottaz (1). One of them is an environmental engineering company; another, a manufacturer.

Claude has another habit: it cites brands that no longer exist. Dextra (2), CPM Braxis (1), Concrete Solutions (1) were absorbed by CI&T, Accenture and Capgemini, and Claude still recommends them as independent companies.

companyChatGPT
25 answers
Claude
25 answers
Google
24 answers
CI&T565
Stefanini444
DB1224
Objective413
Ilegra331
Zup142
Ateliware123
BRQ311
SoftDesign212
UDS Tecnologia212
DevCore202
NextAge301
Accenture030
GFT111
Luby201
Avanade020
BH Softwares002
Capgemini020
DBServer110
Devio101
The table above shows the 20 most cited companies. The other 66 appear in one or two answers each, and their total count is in section 01.
04 · what was left out

Those AI cites that are not competing for this sale

Some of the names in the answers do not develop custom software: they sell ready-made software, such as ERP, transport systems or customer service platforms. AI cites them because the question touches on the subject ("integrar com o ERP", "sistema de logística"). All were counted the same way and stay out of the table: Softplan (8), TOTVS (7), Matera (4), DOT Digital Group (2), AgroX Tech (1), Benner (1), Blip (1), Cedro Technologies (1), Cropware (1), Esplore Tecnologia (1), Inovatech (1), Integra Sistemas (1), Linx (1), Magic Software (1), Maxys (1), Sankhya (1), Senior Sistemas (1), ServiceLogic (1). Also left out are Beltis (1) and Park Place (1), which do IT infrastructure and equipment maintenance.

Why this page does not have the sources consulted section. Only in 7 of the 74 answers can you see which sites the tool read, and ChatGPT did not show that list in any of them. That is too little to say anything about who gets read.
05 · how we measured

How we measure

  • We asked the 25 questions in 3 tools, once each, in September 2026: ChatGPT, Claude and Google. That is 74 answers. Each question was asked in a clean window, temporary conversation without history, from an IP in Brazil. We saved every screenshot.
  • No question includes a company name. None is too generic either: when the question does not state the segment or mention national reach, ChatGPT answers with companies from the city where the question came from.
  • We count only when AI writes the company’s name in the answer. Site cards, maps and company profiles that appear around the answer stay out.
  • The names in the table came from the answers themselves, and each one was checked on the company's website before being included. We did not start from a ready-made list.
06 · limits

What this study does not say

  • Nothing about which company delivers best. What is measured is what AI models answer, not the quality of any project.
  • Nothing about how much each thing is searched. The 25 questions reflect how people ask, not a sample weighted by search volume.
  • Nothing about what AI answers for you. Answers vary with the preceding conversation, region and day. This is a September 2026 snapshot in a clean window.
07 · the questions

The 25 questions, in full

Original Portuguese research evidence, preserved as recorded.

  1. 01quais as melhores empresas de desenvolvimento de software sob medida com atuacao nacional no brasil
  2. 02ranking das maiores software houses do brasil com atuacao nacional
  3. 03empresa de desenvolvimento de aplicativos com atuacao nacional e bons cases para indicar
  4. 04software house com atuacao nacional para desenvolver sistema sob medida para empresa de medio porte
  5. 05melhores fabricas de software do brasil com atuacao nacional em 2026
  6. 06software house para desenvolver um app de fintech
  7. 07empresa para desenvolver aplicativo de saude que cumpra a lgpd
  8. 08software house especializada em marketplace e e-commerce sob medida
  9. 09empresa para desenvolver sistema sob medida para o agronegocio
  10. 10software house para startup desenvolver o mvp do produto
  11. 11empresa de desenvolvimento de plataforma para edtech
  12. 12fabrica de software para desenvolver sistema de logistica e transporte
  13. 13empresa especializada em modernizar sistema legado com ia
  14. 14empresa para terceirizar um squad de desenvolvimento de software
  15. 15software house que desenvolve agentes de ia e integracao com llm para empresas
  16. 16empresa para migrar sistema legado para a nuvem
  17. 17empresa de desenvolvimento de app que trabalha com projeto fechado e escopo definido
  18. 18software house com experiencia em integracao de erp com sistemas internos
  19. 19empresa para desenvolver app nativo para ios e android
  20. 20empresa de alocacao de desenvolvedores seniores para projeto de software
  21. 21quanto custa desenvolver um aplicativo sob medida em 2026
  22. 22como escolher uma software house para desenvolver o sistema da empresa
  23. 23vale mais a pena contratar software house ou montar time interno de desenvolvimento
  24. 24quanto tempo leva para desenvolver um app do zero
  25. 25o que precisa constar no contrato com uma empresa de desenvolvimento de software