Request the four-integrator GenAI delivery bench pack (126 mapped current / 42 mapped jobs)
A work email is all it takes. The pack carries the four-vendor table, the bench-to-book column and the opened official-title list, on the same methodology. Aggregate only; it carries no personal information.
126 seats sit on the bench; 42 sit in the requisition book
The 126 on the bench are delivery stock the four vendors can put on a statement of work today; the 42 in the book are the seats they still have to buy.
126 adds only four rows: Accenture 32, Infosys 23, Cognizant 39, Capgemini 32. It reads opened GenAI delivery titles, not any vendor's company size.
42 adds mapped open jobs on the same titles only: Accenture 1, Infosys 25, Cognizant 2, Capgemini 14. A buyer reads it as how much of the seat the vendor is still short.
Infosys maps 23 current and 25 open jobs. It is the only one of the four rows where the book is thicker than the bench, so that delivery capacity is being sold and built at the same time.
Cognizant maps 39 current, the thickest of the four rows, and 2 open jobs, the thinnest. The largest stock is not the largest buyer, and a buyer has to ask those two questions separately.
Infosys is the hiring wall: 23 on the bench, 25 in the book
Infosys maps 23 current against 25 open jobs. Its bench-to-book ratio is 0.9, the only one of the four below 1.
What a ratio below 1 means
The 25 open jobs outnumber the 23 people already on the seat. A buyer is looking at delivery capacity being expanded, not capacity already built: part of the promised seat has no name on it yet.
The one question to ask in DD
The question is not how many GenAI people the vendor has. It is whether the seats staffed on your programme come from the 23 already on the bench or from the 25 still in the book. Those two answers carry very different delivery risk.
What this does not say
A thick book is not the same as weak delivery; demand may genuinely be rising. It only says the ready-now share of this capacity is low, which is worth turning into a named-commitment clause at bid time.
How this seat was lit up
The Infosys row rests on two opened official titles: GenAI Engineer (Domain Delivery) and Gen/Agentic AI Engineer. It was not padded out with consulting roles.
Cognizant is stock-heavy: 39 on the bench, 2 in the book
Cognizant maps 39 current, the thickest bench of the four rows, against 2 open jobs, the thinnest book. The ratio is 19.5.
The most ready-now capacity
The 39 people already carrying this title are the largest ready-now pool of the four rows. For a programme that has to start this quarter, that is the most direct positive signal on the page.
Two ways to read a book of 2
One reading is that the capacity is already built and needs no more buying. The other is that it is not on the expansion plan. Those diverge over a contract term, so ask which one it is.
Stock is not availability
The 39 is mapped current, not unassigned. What a buyer needs is how much of the 39 is movable on the contract start date, and only the vendor can attest to that layer.
How this seat was lit up
The Cognizant row rests on one opened official title: Python Gen AI Engineer. One title carrying 39 current seats says this vendor names the seat in a very concentrated way.
Nine opened official titles are the map; generic consultant is not
This bench admits opened official titles only. A seat whose title names GenAI / Agentic AI engineering or development is in; a generic consulting role is not.
| Integrator | Opened official titles on this seat |
|---|---|
| Accenture | Senior GenAI Engineer; GenAI Engineer (Mid) (AI&Data); Fullstack AI Engineer (GEN AI) |
| Infosys | GenAI Engineer (Domain Delivery); Gen/Agentic AI Engineer |
| Cognizant | Python Gen AI Engineer |
| Capgemini | GenAI / Agentic AI Developer; Python & GenAI Engineer; Gen AI Engineer |
The four-vendor row-for-row compare a buyer can lift into a DD pack
Four rows on one seat, one as-of and one set of titles. The bench-to-book column is mapped current divided by mapped open jobs — division only, no new figures.
| Integrator | Mapped current | Mapped open jobs | Bench ÷ book | Procurement read |
|---|---|---|---|---|
| Accenture | 32 | 1 | 32.0 | stock on the bench, book almost closed |
| Infosys | 23 | 25 | 0.9 | hiring wall: capacity sold while it is built |
| Cognizant | 39 | 2 | 19.5 | stock-heavy: thickest bench, thinnest book |
| Capgemini | 32 | 14 | 2.3 | the one row where bench and book both move |
| Four-row sum | 126 | 42 | 3.0 | these four rows only, nobody else |
Take this table into your vendor DD: request this pack, or recut the seat on your own shortlist
The pack ships the four rows above, the bench-to-book column and the nine opened official titles, on the same methodology and the same as-of. Swap in the integrators you are actually evaluating and the same seat is recut. A work email is all that is needed.
Where this seat concentrates: the United States, India, the EU
The four official pages opened for this check do not all sit where the vendor brand sits: the Accenture req is Warsaw, the Cognizant req is India, the Capgemini req is New York, and the Infosys req is posted on a global careers board.
United States
Onshore delivery and client-site work are the most expensive line on a rate card. The Capgemini GenAI / Agentic AI Developer req is posted in New York, so an onshore seat is not one vendor's exclusive trait.
India
India is normally where an integrator restocks this seat fastest. The Cognizant Python Gen AI Engineer req sits on its India site, and that should be read together with its thickest bench.
European Union
Data residency and works councils change a delivery plan directly. The Accenture req opened for this check is in Warsaw, inside the EU, which is not the same thing as negotiating a vendor as if it were one location.
This page publishes no country or city split integers. Concentration appears only as a direction for the onshore-risk question: onshore is a per-requisition question, not a per-vendor one. A country-level cut can be rerun on the same methodology inside the vendor pack.
TCS and IBM Consulting are not rows on this bench
Four rows are four rows. Two names a buyer would expect to see are left off the bench, and the reason is written here instead of being filled in with an estimate.
TCS
No titled individual-contributor requisition was opened on this seat. No title means no admission, so TCS gets no row and no other job family is substituted for it.
IBM Consulting
It shares one parent record with Research and Cloud. Unioning those would turn a delivery-bench table into a company table, so they are neither unioned nor split out here.
Also not on this page
No total workforce, no ranking by company size, no personal names. This page reads one seat and compares only on that seat.
Judgments this page does not make
It does not score delivery quality, rank vendors or forecast. It supplies one checkable bench baseline and leaves the rest to a buyer's evaluation sheet.
How the seat was cut, and what the four numbers do not say
The seat cut
A seat enters only when its title names GenAI / Agentic AI or Data-and-AI delivery engineering or development, and only from opened official titles: Accenture Senior GenAI Engineer, GenAI Engineer (Mid) (AI&Data), Fullstack AI Engineer (GEN AI); Infosys GenAI Engineer (Domain Delivery), Gen/Agentic AI Engineer; Cognizant Python Gen AI Engineer; Capgemini GenAI / Agentic AI Developer, Python & GenAI Engineer, Gen AI Engineer. Generic consultant, GenAI SWE and AI Delivery SWE were not used as the map.
Current and open jobs
Mapped current means profiles Metix maps as currently employed at that integrator with a title from the list above. Mapped open jobs means visible open requisitions mapped to the same titles. Neither is official headcount, and neither is a vendor's total workforce.
Names and sums
Each integrator is read on one vendor record; leftover records are not unioned. The 126 / 42 sums add the Accenture, Infosys, Cognizant and Capgemini rows only. The bench-to-book column is this page's own two columns divided and introduces no new figure. This page publishes no country or city split integers.
As-of and the opened official pages
As of August 31, 2026 PT (8/31). Source: Metix AI. Official hiring pages opened for this check: Accenture https://accenture.wd103.myworkdayjobs.com/en-US/AccentureCareers/job/Warsaw-Sienna-39/Senior-Agentic-AI---Data-Engineer_R00318995 ; Infosys https://digitalcareers.infosys.com/global-careers/company-job/description/reqid/151780BR ; Cognizant https://careers.cognizant.com/india-en/jobs/00069794036/python-gen-ai-engineer/ ; Capgemini https://careers.capgemini.com/job/New-York-GenAI-Agentic-AI-Developer/1412574133/ . This page sits in the talent-intelligence report library.
Questions this report answers
- How many GenAI delivery seats do the four system integrators map?
- As of August 31, 2026 PT, Accenture 32, Infosys 23, Cognizant 39 and Capgemini 32 add to 126 mapped current GenAI / Data-and-AI delivery seats, against 42 mapped open jobs on the same titles. These are Metix-mapped counts on opened official titles, not official headcount and not a vendor's total workforce. Source: Metix AI.
- Which integrator is the hiring wall, and which one is stock-heavy?
- Infosys is the hiring wall: 23 mapped current against 25 mapped open jobs, a bench-to-book ratio of 0.9, so the seat is being bought roughly as fast as it is staffed. Cognizant is stock-heavy: 39 mapped current against 2 mapped open jobs, a ratio of 19.5. Accenture sits at 32 current / 1 job and Capgemini at 32 current / 14 jobs. Source: Metix AI.
- Which titles carry this seat, and which vendors are not rows on this bench?
- Nine opened official titles carry the seat: Accenture Senior GenAI Engineer, GenAI Engineer (Mid) (AI&Data), Fullstack AI Engineer (GEN AI); Infosys GenAI Engineer (Domain Delivery), Gen/Agentic AI Engineer; Cognizant Python Gen AI Engineer; Capgemini GenAI / Agentic AI Developer, Python & GenAI Engineer, Gen AI Engineer. Generic consultant, GenAI SWE and AI Delivery SWE were not used as the map. TCS is not a row because no titled individual-contributor req was opened on this seat, and IBM Consulting is not a row because it shares a parent record with Research and Cloud and was not unioned. Official hiring pages opened for this check: https://accenture.wd103.myworkdayjobs.com/en-US/AccentureCareers/job/Warsaw-Sienna-39/Senior-Agentic-AI---Data-Engineer_R00318995 ; https://digitalcareers.infosys.com/global-careers/company-job/description/reqid/151780BR ; https://careers.cognizant.com/india-en/jobs/00069794036/python-gen-ai-engineer/ ; https://careers.capgemini.com/job/New-York-GenAI-Agentic-AI-Developer/1412574133/ .
- What can a procurement team do with this exhibit, and what is in the vendor pack?
- The four rows are a bench baseline for vendor due diligence: read mapped current as delivery stock a vendor can move onto a statement of work today, and mapped open jobs as the seat the vendor still has to buy. The vendor pack ships the same four-vendor table, the bench-to-book column and the opened title list on the same methodology, and the seat can be recut on a buyer's own shortlist. The pack is aggregate only and carries no personal information.
- How should this four-vendor GenAI delivery bench be cited?
- Metix AI Talent Intelligence, 2026-08-31. SI GenAI Delivery Bench 2026: 126 Mapped Seats Against 42 Jobs at Four Integrators | Metix AI. https://metix.ai/reports/mapping/si-genai-delivery-bench-2026
Request the four-integrator GenAI delivery bench pack
The same four-vendor table, the same methodology and the same as-of. Leave a work email and the Metix AI team follows up with the pack. If you would rather cut the seat yourself, the trial below runs it on the three integrators you are evaluating.
Start a Mira trial