<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" version="2.0">
  <channel>
    <title>blog</title>
    <link>https://avarr.ai/insights</link>
    <description>Perspectives on execution intelligence, governed AI and the decisions trading desks make across capital markets, from the team building Avarrai.</description>
    <language>en</language>
    <pubDate>Fri, 20 Nov 2026 23:30:24 GMT</pubDate>
    <dc:date>2026-11-20T23:30:24Z</dc:date>
    <dc:language>en</dc:language>
    <item>
      <title>Models may change. The control boundary does not.</title>
      <link>https://avarr.ai/insights/models-may-change</link>
      <description>&lt;div class="hs-featured-image-wrapper"&gt; 
 &lt;a href="https://avarr.ai/insights/models-may-change" title="" class="hs-featured-image-link"&gt; &lt;img src="https://avarr.ai/hubfs/models-may-change-the-control-boundary-does-not.png" alt="A trader rests his hand on a closed orange gate in a glowing intelligence layer, while teal streams flow only through the open channels." class="hs-featured-image" style="width:auto !important; max-width:50%; float:left; margin:0 15px 15px 0;"&gt; &lt;/a&gt; 
&lt;/div&gt; 
&lt;p&gt;AI has arrived on the trading desk, and heads of trading are asking vendors sharper questions than they were a year ago. Which models sit behind the tool? What data do they use? And when something goes wrong, who is responsible for the trade?&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;AI has arrived on the trading desk, and heads of trading are asking vendors sharper questions than they were a year ago. Which models sit behind the tool? What data do they use? And when something goes wrong, who is responsible for the trade?&lt;/p&gt;  
&lt;p&gt;At FILS Europe this autumn, one head of fixed income trading put the last point very simply. At the end of the day, the trader is responsible for the trade. &lt;a href="https://www.greenwich.com/blog/european-fixed-income-market-structure-focus"&gt;Coalition Greenwich&lt;/a&gt; hears the same from desks across Europe. They welcome AI that removes manual analysis, but they are cautious where it starts to replace judgement, and they insist on being able to challenge what it produces.&lt;/p&gt; 
&lt;p&gt;Those are the right questions. The answers they get are often the wrong ones.&lt;/p&gt; 
&lt;h2&gt;The model is the part that will change&lt;/h2&gt; 
&lt;p&gt;Ask most AI vendors how their product can be trusted and the answer starts with the model. It is the latest one, it has been fine-tuned, and it performs well on a benchmark.&lt;/p&gt; 
&lt;p&gt;The trouble is that the model is the least permanent part of the system. The models available today will be replaced within months, and then replaced again. Any vendor that is serious about quality will keep upgrading.&lt;/p&gt; 
&lt;p&gt;So if a desk's confidence in an AI tool rests on the model, that confidence has to be rebuilt every time the model changes. The same goes for any review, sign-off or oversight that was tied to it. That is not a sound foundation for anything that touches execution.&lt;/p&gt; 
&lt;h2&gt;What should stay fixed is the boundary&lt;/h2&gt; 
&lt;p&gt;A better question is not which model a tool uses, but what that model is allowed to do.&lt;/p&gt; 
&lt;p&gt;In a well-governed system, the boundary around the model matters more than the model itself. It sets out what AI may do, such as interpret, rank, recommend and explain. It sets out what it may not do without a person deciding. It also determines what has to be recorded every time, so that the decision can be reviewed later.&lt;/p&gt; 
&lt;p&gt;When that boundary is designed properly, a model can be upgraded without the desk having to re-learn what the system is permitted to do. The controls, the approvals and the audit trail stay where they were. Only the quality of the advice improves.&lt;/p&gt; 
&lt;p&gt;That is why we put it the way we do. Models may change. The control boundary does not.&lt;/p&gt; 
&lt;h2&gt;Responsibility is a design requirement&lt;/h2&gt; 
&lt;p&gt;"You're responsible for the trade" is often heard as a disclaimer. It should be read as a specification.&lt;/p&gt; 
&lt;p&gt;If the trader carries the responsibility, the system has to make that responsibility possible to exercise. The trader needs to see why a recommendation was made, in terms they recognise. They need to be able to disagree with it, easily and without friction. And there needs to be a record showing what was suggested, what was decided and by whom.&lt;/p&gt; 
&lt;p&gt;An AI tool that cannot do those things is asking the trader to accept accountability for reasoning they cannot see. Few experienced traders will do that, and they are right not to.&lt;/p&gt; 
&lt;h2&gt;An explanation has to be evidence&lt;/h2&gt; 
&lt;p&gt;There is a subtler risk too. Language models are very good at producing explanations that sound convincing. But an explanation the model has composed for itself is not evidence. It is a story.&lt;/p&gt; 
&lt;p&gt;For an explanation to be worth anything on a trading desk, it has to be tied to the facts the recommendation was actually based on. A trader, a compliance officer or a portfolio manager should be able to follow it back to those facts and check it.&lt;/p&gt; 
&lt;h2&gt;Five questions worth asking any AI vendor&lt;/h2&gt; 
&lt;p&gt;For desks evaluating AI tools for execution, these questions tend to separate governed systems from impressive demonstrations.&lt;/p&gt; 
&lt;ul&gt; 
 &lt;li&gt;What exactly is the AI allowed to do, and what is it not allowed to do?&lt;/li&gt; 
 &lt;li&gt;If you change the model tomorrow, what changes for my desk?&lt;/li&gt; 
 &lt;li&gt;Can my traders see why a recommendation was made, and override it easily?&lt;/li&gt; 
 &lt;li&gt;Is there a record of what was recommended, what was decided and by whom?&lt;/li&gt; 
 &lt;li&gt;Where does my data go, and who else benefits from it?&lt;/li&gt; 
&lt;/ul&gt; 
&lt;p&gt;A vendor with clear, specific answers has thought about governance from the start. A vendor whose answers all lead back to the model probably has not.&lt;/p&gt; 
&lt;h2&gt;Governance is part of the infrastructure&lt;/h2&gt; 
&lt;p&gt;In capital markets, intelligence cannot simply be capable. It has to be explainable, controlled and accountable. That cannot be added at the end as a compliance layer. It has to be part of how the system is built.&lt;/p&gt; 
&lt;p&gt;That is the principle behind &lt;a href="https://avarr.ai/platform#governed_ai"&gt;Avarrai's approach to governed AI&lt;/a&gt;. The institution sets the limits, the platform works within them, and the boundary holds whichever model sits inside it.&lt;/p&gt; 
&lt;p&gt;We're looking for a small number of buy-side desks to work with as design partners. If these are the questions your desk is asking, &lt;a href="https://avarr.ai/contact" style="color: #b86130; text-decoration: underline;"&gt;&lt;span style="color: #b86130;"&gt;we'd like to hear from you&lt;/span&gt;&lt;/a&gt;.&lt;/p&gt;  
&lt;img src="https://track-eu1.hubspot.com/__ptq.gif?a=146489653&amp;amp;k=14&amp;amp;r=https%3A%2F%2Favarr.ai%2Finsights%2Fmodels-may-change&amp;amp;bu=https%253A%252F%252Favarr.ai%252Finsights&amp;amp;bvt=rss" alt="" width="1" height="1" style="min-height:1px!important;width:1px!important;border-width:0!important;margin-top:0!important;margin-bottom:0!important;margin-right:0!important;margin-left:0!important;padding-top:0!important;padding-bottom:0!important;padding-right:0!important;padding-left:0!important; "&gt;</content:encoded>
      <category>Execution intelligence</category>
      <category>Governed AI</category>
      <pubDate>Mon, 09 Nov 2026 19:46:13 GMT</pubDate>
      <author>abdullah@hiyatt.com (Abdullah Hiyatt)</author>
      <guid>https://avarr.ai/insights/models-may-change</guid>
      <dc:date>2026-11-09T19:46:13Z</dc:date>
    </item>
    <item>
      <title>The tape shows you everything. Then what?</title>
      <link>https://avarr.ai/insights/the-tape-shows-you-everything</link>
      <description>&lt;div class="hs-featured-image-wrapper"&gt; 
 &lt;a href="https://avarr.ai/insights/the-tape-shows-you-everything" title="" class="hs-featured-image-link"&gt; &lt;img src="https://avarr.ai/hubfs/the-tape-shows-you-everything.png" alt="A flood of teal market data rises from a city into a glass intelligence layer, where a trader stands at the start of one clear orange path." class="hs-featured-image" style="width:auto !important; max-width:50%; float:left; margin:0 15px 15px 0;"&gt; &lt;/a&gt; 
&lt;/div&gt; 
&lt;p&gt;For years the standing complaint in fixed income was that you couldn't see the market. Prints were scattered, delayed or simply unavailable. Every desk built its own partial picture from dealer runs, composites and whatever else it could piece together.&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;For years the standing complaint in fixed income was that you couldn't see the market. Prints were scattered, delayed or simply unavailable. Every desk built its own partial picture from dealer runs, composites and whatever else it could piece together.&lt;/p&gt; 
&lt;p&gt;Regulators have answered that complaint. The UK now has a consolidated tape for bonds, the EU has appointed its own provider, and the US has had TRACE for years. Reporting volumes are up sharply, and the tape is available to anyone who wants it.&lt;/p&gt; 
&lt;p&gt;Yet a large share of buy-side traders aren't using it.&lt;/p&gt; 
&lt;h2&gt;The tape is live, but most desks aren't looking at it&lt;/h2&gt; 
&lt;p&gt;At FILS Europe in October, a third of buy-side attendees said they &lt;a href="https://www.fi-desk.com/almost-half-buy-side-traders-not-using-uk-bond-tape/"&gt;had no plans to use UK tape data at all&lt;/a&gt;. Only a small minority of those with access use it day to day. &lt;a href="https://www.greenwich.com/blog/european-fixed-income-market-structure-focus"&gt;Coalition Greenwich&lt;/a&gt; reached a similar conclusion. The tape is widely seen as important, yet few desks have it in front of their traders.&lt;/p&gt; 
&lt;p&gt;The reasons traders give are telling. They need data that tells them something they don't already know. Raw prints need cleaning before they're useful, and the tape providers themselves say &lt;a href="https://www.fi-desk.com/fils-europe-bond-ctps-fight-against-free-riders-and-pollution/"&gt;deeper cleaning and analytics fall outside their remit&lt;/a&gt;. Much of the market also remains invisible when it matters most. By one estimate, &lt;a href="https://www.fi-desk.com/fca-defends-80-ctp-deferrals-as-helping-market-makers-under-stress/"&gt;less than a fifth of UK corporate bond volume was reported in real time&lt;/a&gt; earlier this year, with a substantial share deferred for weeks or even months.&lt;/p&gt; 
&lt;p&gt;None of this makes the tape a failure. It is a genuine step forward, and it will improve. What it does show is that access was never the real bottleneck.&lt;/p&gt; 
&lt;h2&gt;From access to relevance&lt;/h2&gt; 
&lt;p&gt;The tape is only the newest source in an already crowded picture. No trader can watch every print in every bond they might care about, any more than they can read every axe, IOI and dealer run that lands on their screens or in their mailbox each morning. Axes in particular remain notoriously noisy, and the market has lived with that for years.&lt;/p&gt; 
&lt;p&gt;So the industry is turning to the next problem. A new generation of tools focuses on relevance. They read what a desk holds, check prints, axes, IOIs and inventory against it, and surface the handful of things that matter, with a short explanation of why.&lt;/p&gt; 
&lt;p&gt;That is useful. It turns a firehose into a shortlist, and it reflects what traders say they want. According to Coalition Greenwich, traders now care more about the quality and usability of data than about having more of it.&lt;/p&gt; 
&lt;p&gt;But a relevant alert is still only an alert.&lt;/p&gt; 
&lt;h2&gt;The next ten trades&lt;/h2&gt; 
&lt;p&gt;A useful test for any new tool is a simple one. Would it materially change the outcome of the desk's next ten trades?&lt;/p&gt; 
&lt;p&gt;It is a good test, and it explains the tape's slow adoption better than any survey. Data only changes an outcome when it changes a decision. When something relevant appears, the trader still has to decide whether to act now or wait. If the answer is to act, they must choose the protocol, the venues, the dealers, the size and the sequence. They must judge what counts as a good fill and when the plan should change. Sometimes the right answer is to do nothing at all.&lt;/p&gt; 
&lt;p&gt;The decision then has to be carried through across RFQ platforms, streams, chat and voice, without giving too much away. Afterwards, someone needs to know what was decided, what happened and why.&lt;/p&gt; 
&lt;p&gt;The market has invested heavily at both ends of this. Data, analytics and alerts sit at one end, with venues, protocols and execution platforms at the other. The middle, where the decision is made and connected to what follows, still lives mostly in the trader's head, a chat window and a spreadsheet.&lt;/p&gt; 
&lt;h2&gt;Why the middle matters more now&lt;/h2&gt; 
&lt;p&gt;Better data doesn't lighten the load on that middle step. It adds to it. Every new signal is one more thing a trader is expected to weigh, quickly and on a busy day, across more line items than any one person can hold in their head.&lt;/p&gt; 
&lt;p&gt;AI raises the stakes further. Desks welcome AI that takes away manual analysis, but they are cautious where it starts to replace judgement, and they insist on being able to challenge what it produces. As these tools move from flagging what matters to suggesting what to do, and eventually to acting, the questions become sharper. Who made the decision? What evidence was it based on? Within what limits?&lt;/p&gt; 
&lt;p&gt;Expectations of evidence are rising too. Portfolio managers, best execution reviews and oversight committees increasingly want to understand not just the price achieved, but the reasoning behind how an order was worked.&lt;/p&gt; 
&lt;h2&gt;What the missing middle needs&lt;/h2&gt; 
&lt;p&gt;Closing the gap doesn't mean replacing the systems desks already rely on. The tools that get adopted are the ones that fit the way desks already work. What desks need is help at the moment of decision, not more information before it. That help has to leave the trader firmly in charge, work with the channels the desk already trusts, and leave a clear record of what was decided and why.&lt;/p&gt; 
&lt;p&gt;The trader stays in charge throughout. The difference is that the decision is no longer invisible.&lt;/p&gt; 
&lt;h2&gt;Seeing is the start, not the finish&lt;/h2&gt; 
&lt;p&gt;The consolidated tape is a real step forward for fixed income, and the tools that make sense of it will add to its value. But seeing the market was never the job. The job is deciding what to do about it, and doing it well.&lt;/p&gt; 
&lt;p&gt;That is the layer Avarrai is building. It is execution intelligence that sits between what the desk can see and what it does, with the trader in control and the evidence captured.&lt;/p&gt; 
&lt;p&gt;We're looking for a small number of buy-side desks to work with as design partners. If this gap sounds familiar, &lt;a href="https://avarr.ai/contact" style="color: #B86130; text-decoration: underline;"&gt;&lt;span style="color: #B86130;"&gt;we'd like to hear from you&lt;/span&gt;&lt;/a&gt;.&lt;/p&gt;  
&lt;img src="https://track-eu1.hubspot.com/__ptq.gif?a=146489653&amp;amp;k=14&amp;amp;r=https%3A%2F%2Favarr.ai%2Finsights%2Fthe-tape-shows-you-everything&amp;amp;bu=https%253A%252F%252Favarr.ai%252Finsights&amp;amp;bvt=rss" alt="" width="1" height="1" style="min-height:1px!important;width:1px!important;border-width:0!important;margin-top:0!important;margin-bottom:0!important;margin-right:0!important;margin-left:0!important;padding-top:0!important;padding-bottom:0!important;padding-right:0!important;padding-left:0!important; "&gt;</content:encoded>
      <category>Execution intelligence</category>
      <category>Market structure</category>
      <pubDate>Mon, 09 Nov 2026 19:02:40 GMT</pubDate>
      <author>abdullah@hiyatt.com (Abdullah Hiyatt)</author>
      <guid>https://avarr.ai/insights/the-tape-shows-you-everything</guid>
      <dc:date>2026-11-09T19:02:40Z</dc:date>
    </item>
    <item>
      <title>Not trading is a decision. Nobody records it.</title>
      <link>https://avarr.ai/insights/not-trading-is-a-decision</link>
      <description>&lt;div class="hs-featured-image-wrapper"&gt; 
 &lt;a href="https://avarr.ai/insights/not-trading-is-a-decision" title="" class="hs-featured-image-link"&gt; &lt;img src="https://avarr.ai/hubfs/not-trading-is-a-decision-3.png" alt="A trader stands on a glowing intelligence layer beside an orange stream of light held at its surface, above a teal city of data." class="hs-featured-image" style="width:auto !important; max-width:50%; float:left; margin:0 15px 15px 0;"&gt; &lt;/a&gt; 
&lt;/div&gt; 
&lt;p&gt;Every experienced fixed income trader knows that some of the best decisions of the day are the trades they didn't do. The order that arrived as liquidity was drying up. The bond already shown to three dealers that morning. The size that would have moved the price against the fund just ahead of month-end.&lt;/p&gt;</description>
      <content:encoded>&lt;p&gt;Every experienced fixed income trader knows that some of the best decisions of the day are the trades they didn't do. The order that arrived as liquidity was drying up. The bond already shown to three dealers that morning. The size that would have moved the price against the fund just ahead of month-end.&lt;/p&gt;  
&lt;p&gt;Holding back, working an order more patiently or going back to the portfolio manager is part of the job. Nobody needs software to tell a good trader that.&lt;/p&gt; 
&lt;p&gt;What's missing is something else. The decision not to trade, or not yet, leaves almost no trace.&lt;/p&gt; 
&lt;h2&gt;What gets measured is what got executed&lt;/h2&gt; 
&lt;p&gt;Transaction cost analysis and best execution reviews are built around fills. They measure the price achieved against arrival, against a composite and against cover. That is useful, but it only covers the trades that happened.&lt;/p&gt; 
&lt;p&gt;An order that was held for two days, worked in smaller pieces or handed back to the PM produces no record of the reasoning, no evidence of the market at the time and no measured outcome. If that patience saved the fund twenty basis points, nobody can show it. If it cost the fund, nobody learns from that either.&lt;/p&gt; 
&lt;h2&gt;Where that costs desks&lt;/h2&gt; 
&lt;p&gt;&lt;strong&gt;The PM conversation.&lt;/strong&gt; When a trader pushes back on an order, it usually comes down to experience against urgency. A trader with a strong record tends to win that argument, and a newer one often doesn't. Either way, the judgement is never written down.&lt;/p&gt; 
&lt;p&gt;&lt;strong&gt;Scale.&lt;/strong&gt; A senior trader can hold the context for a handful of difficult bonds. Spread that judgement across a 200-line list, a volatile session or a colleague covering the book, and it thins out quickly.&lt;/p&gt; 
&lt;p&gt;&lt;strong&gt;Consistency.&lt;/strong&gt; Two traders on the same desk can make opposite calls on the same order on the same day. Without a record, nobody sees it, and the desk can't settle which approach works.&lt;/p&gt; 
&lt;p&gt;&lt;strong&gt;Oversight.&lt;/strong&gt; Compliance and best execution committees review what was executed. Decisions not to execute can matter just as much to client outcomes, but they rarely come into view.&lt;/p&gt; 
&lt;h2&gt;What changes when "not yet" is recorded&lt;/h2&gt; 
&lt;p&gt;Suppose the decision to wait, stage or escalate is captured at the moment it is made. Several things change.&lt;/p&gt; 
&lt;p&gt;The PM conversation becomes a discussion of evidence rather than a contest of opinions. Oversight extends to decisions, not only fills. And judgement that used to live in one trader's head becomes part of the desk's institutional memory, which doesn't walk out of the door when a trader moves on.&lt;/p&gt; 
&lt;h2&gt;The trader still decides&lt;/h2&gt; 
&lt;p&gt;None of this replaces judgement. The decision stays with the trader, and accountability stays with the desk. The point is to make good judgement visible, consistent and measurable, and to give the desk a way to prove it.&lt;/p&gt; 
&lt;p&gt;That is part of what Avarrai is building. It is an execution intelligence layer that treats waiting, escalating and not trading today as decisions in their own right, recorded with the same care as an executed trade.&lt;/p&gt; 
&lt;p&gt;We're looking for a small number of buy-side fixed income desks to work with as design partners. If your desk recognises this problem, &lt;a href="https://avarr.ai/contact" style="color: #b86130; text-decoration: underline;"&gt;&lt;span style="color: #b86130;"&gt;we'd like to hear from you&lt;/span&gt;&lt;/a&gt;.&lt;/p&gt;  
&lt;img src="https://track-eu1.hubspot.com/__ptq.gif?a=146489653&amp;amp;k=14&amp;amp;r=https%3A%2F%2Favarr.ai%2Finsights%2Fnot-trading-is-a-decision&amp;amp;bu=https%253A%252F%252Favarr.ai%252Finsights&amp;amp;bvt=rss" alt="" width="1" height="1" style="min-height:1px!important;width:1px!important;border-width:0!important;margin-top:0!important;margin-bottom:0!important;margin-right:0!important;margin-left:0!important;padding-top:0!important;padding-bottom:0!important;padding-right:0!important;padding-left:0!important; "&gt;</content:encoded>
      <category>Fixed income</category>
      <category>Execution intelligence</category>
      <pubDate>Mon, 09 Nov 2026 10:56:05 GMT</pubDate>
      <author>abdullah@hiyatt.com (Abdullah Hiyatt)</author>
      <guid>https://avarr.ai/insights/not-trading-is-a-decision</guid>
      <dc:date>2026-11-09T10:56:05Z</dc:date>
    </item>
  </channel>
</rss>
