{"id":60342,"date":"2026-09-16T00:07:55","date_gmt":"2026-09-16T07:07:55","guid":{"rendered":"https:\/\/svch.io\/silicon-valley-certification-hub-chief-ai-officer-manager-agent-hierarchy-quality\/"},"modified":"2026-09-16T00:07:55","modified_gmt":"2026-09-16T07:07:55","slug":"silicon-valley-certification-hub-chief-ai-officer-manager-agent-hierarchy-quality","status":"publish","type":"post","link":"https:\/\/svch.io\/es\/silicon-valley-certification-hub-chief-ai-officer-manager-agent-hierarchy-quality\/","title":{"rendered":"The Manager Agent Is Costing You Quality — Silicon Valley Certification Hub Chief AI Officer Research"},"content":{"rendered":"
\n
SVCH Research Review — September 2026<\/div>\n

The Manager Agent Is Costing You Quality — Silicon Valley Certification Hub Chief AI Officer Research<\/h1>\n
\n 📄 arXiv: 2609.14767<\/span>
\n 🏢 Leiden University<\/span>
\n 📅 September 2026<\/span>\n <\/div>\n
Researchers:<\/strong> Burak Agachan · Max van Duijn · Amirhossein Zohrehvand<\/div>\n<\/div>\n

43 paired business reports. 86 runs. The only thing that changed between them was whether a manager could send the work back.<\/p>\n

The team without that power scored higher. Reports were judged better on utility and clearer to read. They hedged 53% less. And they did it on 51.5% fewer tokens. I read the abstract twice because I assumed I had it backwards.<\/p>\n

Almost every multi-agent framework you can buy ships with a manager agent. A supervisor that reviews what the workers produced and, if it does not like it, orders a revision. That is the default org chart for AI teams. This paper is the first clean test of just that one link on open-ended business work, and it says the default is making your output worse.<\/p>\n

\"Silicon<\/p>\n

\n
+0.42<\/div>\n
The flat team scored higher on utility, with a statistically solid effect (d = 0.42, p = 0.009).<\/div>\n
Methodology from over a century of organizational theory predicted the opposite. It also cost 51.5% more tokens to run the version with a boss.<\/div>\n<\/div>\n

Why This Paper Matters<\/h2>\n

Every enterprise now has the same quiet argument happening inside it. One camp wants a supervisor layer: a manager agent that checks the team’s work before anything ships. It feels responsible. It looks like governance. The other camp wants speed and does not want to fund an extra tier of token spend.<\/p>\n

Until now, neither side had evidence. Prior comparisons swapped entire frameworks on math and coding tasks where you can check the answer. That tests everything at once and tells you nothing about the specific managerial power to reject and force rework. On open-ended work, like a business intelligence report, nobody had isolated it.<\/p>\n

The question matters because the answer is expensive either way. A supervisory tier is not free. This paper prices it. And the price is not just the tokens, it is the quality of the work coming out the other end.<\/p>\n

Methodology, Explained Simply<\/h2>\n

The design is the most impressive part, and it is simple to describe. Take five AI agents working as a team on a business intelligence reporting task. Same five agents. Same roles. Same prompts. Same tools. Same models. Same data. Same shared workspace. Now run it 86 times on 43 pairs of products.<\/p>\n

Change exactly one thing: in the hierarchical version, the manager has the authority to reject a worker’s output and demand a revision. In the flat version, it does not. That is the whole experiment. One link in the org chart.<\/p>\n

Then score every report twice over. A panel of five different AI models judges each report across six dimensions, grouped into two headline scores: Writing Clarity and Utility. Separately, a deterministic check verifies the report against the specification, the hard factual requirements. So you get both a judgment call on quality and a hard pass or fail on whether the team actually delivered what was asked.<\/p>\n

They also ran the robustness checks you would want. Leave-one-judge-out re-estimation, to make sure no single judging model was driving the result. And they verified the reports were the same length, so the flat team was not winning simply by writing more or less.<\/p>\n

\"Silicon<\/p>\n

Results and Practical Insights<\/h2>\n

The flat organization won on both headline dimensions. Utility came in at d = 0.42, p = 0.009. Writing Clarity at d = 0.34, p = 0.030. Not huge effects, but consistent and statistically sound across a paired design.<\/p>\n

\n
\n
4.70<\/div>\n
Flat, Strategic Depth<\/div>\n<\/p><\/div>\n
\n
4.56<\/div>\n
Hierarchical, Strategic Depth<\/div>\n<\/p><\/div>\n
\n
53%<\/div>\n
More hedging, hierarchical<\/div>\n<\/p><\/div>\n
\n
51.5%<\/div>\n
Extra tokens for no gain<\/div>\n<\/p><\/div>\n<\/div>\n

Here is the part that changed how I think about this. The hierarchical writer’s first draft was indistinguishable from the flat team’s finished report. The gap did not exist at the start. It opened inside the revision loop. Every round of forced rework cost 0.14 points of Writing Clarity. With zero loops, clarity sat at 4.48. By three loops, it had fallen to 3.79.<\/p>\n

\n
\n \"Silicon
Writing Clarity falls with each revision loop: 4.48 when none were attempted, 3.79 at three loops.<\/figcaption><\/figure>\n<\/figure>\n

Specification accuracy was at ceiling in both organizations. Both teams hit the hard requirements. So the manager was not catching factual errors on the way. It was rewriting perfectly acceptable work into something more hedged, more cautious, more covered. Every revision made the report safer to defend and less useful to act on.<\/p>\n

The supervisory tier cost 51.5% more tokens. For zero measurable verifiable gain. That is a line item, and it is also a warning: if you cannot tell why the manager is needed, the manager may be the problem.<\/p>\n

\n
Chief AI Officer Certification<\/div>\n

Can your manager agent actually verify the work, or does it just have an opinion about it?<\/h3>\n

At Silicon Valley Certification Hub, we help operations and enterprise AI leaders evaluate and deploy agent teams that fit their actual business processes, including whether a supervisory tier earns its keep.<\/p>\n

Book a Strategy Call →<\/a>\n<\/div>\n

What This Means for Your Chief AI Officer<\/h2>\n

The rule this paper lands on is clean. A supervisor pays for itself when it can verify. It becomes a liability when it can only opine. A verifier checks work against a test, an oracle, an exact specification, something objective. An opinion-holder reads the work and says it could be better. The first catches real errors. The second just spends tokens and adds hedging.<\/p>\n

So the question for any AI Assessment for companies is not “should we have a manager agent.” It is “what exactly is our manager agent doing, and can we prove it helps.” That is a question you can answer before you scale the layer, and it is cheaper to answer now than after the token bill arrives.<\/p>\n

Key Takeaways for Operations and Enterprise AI Leaders<\/h2>\n
\n
\n
1<\/div>\n
Ask what your manager agent can verify.<\/strong> If it holds a test, a spec, or an oracle, the supervisory tier may be worth it. If it only has taste, you are buying hedging at full price. Most production frameworks default to the second kind.<\/div>\n<\/p><\/div>\n
\n
2<\/div>\n
Instrument your revision loops.<\/strong> Each loop cost 0.14 points of clarity in this study. If you do not count loops, you cannot see the damage. Make loop count a monitored metric on any agent pipeline that can force rework.<\/div>\n<\/p><\/div>\n
\n
3<\/div>\n
Watch hedging, not just scores.<\/strong> The hierarchical reports hedged 53% more. Hepding is cheap to generate and expensive on the page. Cap the vague qualifiers in your prompts and measure the change.<\/div>\n<\/p><\/div>\n
\n
4<\/div>\n
Price the extra tier before you ship it.<\/strong> 51.5% more tokens for no quality gain is a real budget line. Run your own paired test on your own task. It is a weekend of compute, not a quarter of consulting.<\/div>\n<\/p><\/div>\n
\n
5<\/div>\n
Decide what the manager adds beyond the org chart.<\/strong> If the first draft is already indistinguishable from the finished report, the supervision is theater. Which of your agent roles is actually earning its tokens?<\/div>\n<\/p><\/div>\n<\/div>\n

Thanks to All Authors<\/h2>\n
\n
\n
\n
BA<\/div>\n
\n
Burak Agachan<\/div>\n
Leiden University, Leiden, Netherlands<\/div>\n<\/div><\/div>\n
\n
MD<\/div>\n
\n
Max van Duijn<\/div>\n
Leiden University, Leiden, Netherlands<\/div>\n<\/div><\/div>\n
\n
AZ<\/div>\n
\n
Amirhossein Zohrehvand<\/div>\n
Leiden University, Leiden, Netherlands<\/div>\n<\/div><\/div>\n<\/p><\/div>\n<\/div>\n
\n

Want to know how this applies to your company?<\/strong><\/p>\n

At Silicon Valley Certification Hub, we help you align AI + Strategy. Our team works directly with your directors and teams to assess AI readiness, identify gaps, and build a clear path forward — tailored to your business context.<\/p>\n

Book a time with our CEO, Alejandro Cuauhtemoc-Mejia:
\nhttps:\/\/calendar.app.google\/2ihQf2JH3D9uJBe68<\/a><\/p>\n

Silicon Valley Certification Hub
\n3000 El Camino Real, Building 4, Palo Alto, CA<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

Silicon Valley Certification Hub reviews the latest AI research for Chief AI Officers. Flat AI agent teams beat hierarchical ones on quality, and cost 51.5% fewer tokens. Here is what that means for your company.<\/p>\n","protected":false},"author":155,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"content-type":"","_monsterinsights_skip_tracking":false,"advanced_seo_description":"","jetpack_seo_html_title":"","jetpack_seo_noindex":false,"jetpack_seo_schema_type":"","_price":"","_stock":"","_tribe_ticket_header":"","_tribe_default_ticket_provider":"","_tribe_ticket_capacity":"","_ticket_start_date":"","_ticket_end_date":"","_tribe_ticket_show_description":"","_tribe_ticket_show_not_going":false,"_tribe_ticket_use_global_stock":"","_tribe_ticket_global_stock_level":"","_global_stock_mode":"","_global_stock_cap":"","_tribe_rsvp_for_event":"","_tribe_ticket_going_count":"","_tribe_ticket_not_going_count":"","_tribe_tickets_list":"[]","_tribe_ticket_has_attendee_info_fields":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_post_was_ever_published":false},"categories":[24],"tags":[770,543,544,551,542,707,742,771,541,480,717],"class_list":["post-60342","post","type-post","status-publish","format-standard","hentry","category-research","tag-ai-agent-teams","tag-ai-assessment","tag-ai-for-executives","tag-ai-governance","tag-chief-ai-officer","tag-enterprise-ai-adoption","tag-multi-agent-systems","tag-operations","tag-silicon-valley-certification-hub","tag-svch","tag-workforce-productivity"],"acf":[],"jetpack_likes_enabled":true,"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/posts\/60342","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/users\/155"}],"replies":[{"embeddable":true,"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/comments?post=60342"}],"version-history":[{"count":0,"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/posts\/60342\/revisions"}],"wp:attachment":[{"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/media?parent=60342"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/categories?post=60342"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/svch.io\/es\/wp-json\/wp\/v2\/tags?post=60342"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}