Attacking the Belly of the Beast: How AI-First Startups Can Solve Supply & Demand Planning

Bivek Adhikari — December 27, 2024

Pic­ture this: It is mid- Novem­ber, and you are man­ag­ing a com­plex glob­al retail sup­ply chain, prepar­ing for the hol­i­day season’s rush. Demand is soar­ing, and the pres­sure is mount­ing. A sin­gle hiccup—whether it is a delayed ship­ment, a sup­pli­er issue, or an unex­pect­ed surge in demand—could send shock­waves through your care­ful­ly coor­di­nat­ed net­work, dis­rupt­ing inven­to­ry and pro­duc­tion sched­ules across regions. Despite hav­ing access to vast amounts of data, your deci­sions are based on out­dat­ed fore­casts and reac­tive process­es that fail to cap­ture the real-time dynam­ics need­ed to nav­i­gate today’s com­plex glob­al sup­ply chain. Unfor­tu­nate­ly, this is still the real­i­ty for many busi­ness­es, as over 50% of the largest U.S. importers con­tin­ue to rely on spread­sheets.

Now, imag­ine an AI-pow­ered solu­tion that not only pre­dicts demand but also antic­i­pates sup­ply-side dis­rup­tions and pro­vides pre­scrip­tive insights—all in real-time. This is the poten­tial of AI-dri­ven sup­ply chain intel­li­gence (SCI), but it remains large­ly unre­al­ized par­tic­u­lar­ly in the crit­i­cal area of demand and sup­ply plan­ning. This core chal­lenge, the “bel­ly of the beast” in sup­ply chain man­age­ment, is where star­tups can make the biggest impact, trans­form­ing an entire indus­try and unlock­ing unprece­dent­ed oper­a­tional resilience.

A vision of an enter­prise archi­tec­ture for demand and sup­ply plan­ning: a sys­tem of AI agents.

Prob­lem: Ser­vice-Heavy Solu­tions Dom­i­nate Sup­ply and Demand Plan­ning

We believe con­sult­ing giants like Accen­ture, McK­in­sey & Com­pa­ny, and Deloitte dom­i­nate today’s sup­ply chain land­scape, offer­ing high-touch, cus­tomized solu­tions to help com­pa­nies man­age inven­to­ry, fore­cast demand shifts, and mit­i­gate risks. While effec­tive, these ser­vices often come at a high cost, involve long time­lines, and may have lim­it­ed  scale. In an era where dis­rup­tions have become more fre­quent and unpre­dictable, this approach is no longer sus­tain­able.

Foun­da­tion Cap­i­tal’s equat­ing of the rise of AI Agents to an evo­lu­tion from sta­t­ic Soft­ware-as-a-Ser­vice to what they term Ser­vice-as-Soft­ware is par­tic­u­lar­ly rel­e­vant in sup­ply chain. This shift brings the respon­si­bil­i­ty for achiev­ing busi­ness out­comes onto the AI sys­tems them­selves, mak­ing them invalu­able in time­ly demand and sup­ply plan­ning. McK­in­sey reports that 90% of IT lead­ers are aim­ing to over­haul their sup­ply chain plan­ning sys­tems. How­ev­er, the mar­ket lacks a tru­ly pro­duc­tized, AI-first solu­tion capa­ble of real-time demand and sup­ply plan­ning at scale. This gap under­scores an enor­mous mar­ket oppor­tu­ni­ty for star­tups to intro­duce scal­able, auto­mat­ed plat­forms that can replace tra­di­tion­al ser­vice-heavy mod­els. The SCI mar­ket has been steadi­ly grow­ing, reach­ing $22 bil­lion in 2023 with expec­ta­tions to hit a 15.4% CAGR through 2026.⁽¹⁾ How­ev­er, demand and sup­ply plan­ning remains dom­i­nat­ed by bespoke con­sult­ing ser­vices, leav­ing a sig­nif­i­cant white space for AI-dri­ven, pro­duc­tized solu­tions.

Why Sup­ply and Demand Plan­ning is the “Bel­ly of the Beast”

Sup­ply and demand plan­ning is one of the most com­plex aspects of sup­ply chain man­age­ment. It requires real-time data inte­gra­tion from mul­ti­ple sources—historical sales data, mar­ket trends, weath­er pat­terns, eco­nom­ic conditions—then trans­lates this data into action­able insights for inven­to­ry and pro­duc­tion sched­ules across glob­al oper­a­tions. Tra­di­tion­al sys­tems, designed more for sta­bil­i­ty than adapt­abil­i­ty, strug­gle with the agili­ty need­ed to man­age this vast data inte­gra­tion.

As Ajay Agar­w­al and Zeeza Cole from Bain Cap­i­tal Ven­tures stat­ed in their the­sis, mov­ing beyond lega­cy Elec­tron­ic Data Inter­change (EDI) systems—which were rev­o­lu­tion­ary in the 1970s but now hin­der agility—is key. AI solu­tions bring the ben­e­fit of adapt­able, real-time insights that are dif­fi­cult to achieve with sta­t­ic, rules-based sys­tems like EDI. McK­in­sey esti­mates that effec­tive demand and sup­ply plan­ning can reduce out-of-stock rates by up to 65% and slash excess inven­to­ry by up to 30%. These sta­tis­tics under­score the impact that auto­mat­ed, AI-pow­ered solu­tions could have on sup­ply chain man­age­ment. Yet, most solu­tions remain tied to con­sult­ing-led mod­els that lim­it scal­a­bil­i­ty, flex­i­bil­i­ty, and afford­abil­i­ty.

Ver­ti­cal spe­cif­ic plan­ning issues, as demon­strat­ed in food waste and sus­tain­abil­i­ty chal­lenges, also demand a more dynam­ic solu­tion. For exam­ple, in the food indus­try alone, sup­ply chain inef­fi­cien­cies con­tribute to over  , with per­isha­bil­i­ty, sea­son­al­i­ty, and trans­porta­tion hur­dles ampli­fy­ing loss­es. Solu­tions that can reduce over­age by accu­rate­ly fore­cast­ing demand and syn­chro­niz­ing sup­ply could sig­nif­i­cant­ly cut waste and enhance sustainability—a grow­ing imper­a­tive across indus­tries.

Sup­ply chain intel­li­gence along with D2C & cus­tomer chan­nel man­age­ment mar­kets add up to a $20B+ tar­get mar­ket size grow­ing at 10%-15% across the board.

The Evo­lu­tion from EDI to AI-Dri­ven Sup­ply Chain Plan­ning

The con­cept of EDI, which emerged in the 1970s, was rev­o­lu­tion­ary for sup­ply chains, enabling com­pa­nies to exchange doc­u­ments elec­tron­i­cal­ly and reduce man­u­al errors. How­ev­er, EDI intro­duced new com­plex­i­ties: indus­tries adopt­ed dif­fer­ent stan­dards, lead­ing to inter­op­er­abil­i­ty issues that per­sist today. As SCI has grown at 11% CAGR over the past decade, mov­ing away from EDI to real-time, cloud-based solu­tions have become essen­tial.

In today’s sup­ply chains, data remains frag­ment­ed across silos—some data is stored on-premis­es, some in cloud environments—with sys­tems oper­at­ing inde­pen­dent­ly. This frag­men­ta­tion leads to fric­tion and inef­fi­cien­cies. To deliv­er real-time fore­cast­ing, com­pa­nies need a uni­fied plat­form that inte­grates these dis­parate data sources. How­ev­er, data is fre­quent­ly incon­sis­tent, unstruc­tured, and frag­ment­ed, mak­ing real-time plan­ning chal­leng­ing.

Talks with Star­tups: A Push Toward Tack­ling the Core

In my con­ver­sa­tions with point-solu­tion star­tups, I have repeat­ed­ly empha­sized the impor­tance of attack­ing the “bel­ly of the beast,” instead of focus­ing on a sin­gle pain point first and build­ing out the solu­tions set from there on. Zip Co, which raised $190M in Octo­ber 2024 at $2.01B pre-mon­ey, is launch­ing a pro­cure­ment orches­tra­tion plat­form to sim­pli­fy the com­plex­i­ties of sup­ply chain man­age­ment. Sim­i­lar­ly, com­pa­nies like Order­ful, Noodle.ai, Chain.io, Crstl, Crisp, Grub­Mar­ket Inc., and Keel­var are tack­ling niche areas like EDI mod­ern­iza­tion, pro­cure­ment automa­tion, inven­to­ry opti­miza­tion, and sup­pli­er risk.

Most star­tups begin by solv­ing niche prob­lems with­in pro­cure­ment, sourc­ing, finance automa­tion, or risk man­age­ment because they are eas­i­er to address and offer quick­er wins. How­ev­er, attack­ing the core problem—demand and sup­ply planning—could unlock far greater val­ue across entire sup­ply chains. Ben­jamin Fels from Pen­du­lum Sys­tems and San­jay Sai­ni from Stem­ly, for exam­ple, are pur­su­ing demand and sup­ply plan­ning more direct­ly. Pen­du­lum is tak­ing an incre­men­tal approach, ini­tial­ly pro­vid­ing ser­vices while set­ting sights on even­tu­al pro­duc­ti­za­tion. This method allows them to build foun­da­tion­al under­stand­ing and capa­bil­i­ties before com­mit­ting to ful­ly auto­mat­ed solu­tions.

Mar­ket map of growth stage incum­bents and emerg­ing play­ers in the SCI space.

Due to these vary­ing approach­es, the white space remains unfilled—no com­pa­ny has yet devel­oped an inte­grat­ed, AI-dri­ven plat­form that address­es real-time demand and sup­ply plan­ning on a glob­al scale. As not­ed by Ajay Agar­w­al and Zeeza Cole from Bain Cap­i­tal Ven­tures, To tru­ly dis­rupt glob­al sup­ply chains, star­tups need to think big­ger and build plat­forms that address end-to-end vis­i­bil­i­ty and deci­sion-mak­ing across all functions—from raw mate­ri­als sourc­ing to the final deliv­ery.

Start Small with Low-Resis­tance Inte­gra­tion Points

For ear­ly-stage star­tups look­ing to break into SCI, tack­ling the data inte­gra­tion chal­lenge in one sweep­ing move can be over­whelm­ing. An effec­tive approach is to focus on low-resis­tance inte­gra­tion points, lay­ing the ground­work for a scal­able foun­da­tion. The key steps include:

Inte­grate with Acces­si­ble Data Sources First: Start with cloud-based tools and open APIs from logis­tics providers. This builds an ini­tial lay­er of acces­si­ble data, cre­at­ing foun­da­tion­al pipelines for future growth.

Map the Data Ecosys­tem Grad­u­al­ly: Devel­op sys­tem dia­grams to map each inte­gra­tion point, cre­at­ing a clear view of data flows across the sup­ply chain.

Auto­mate Data Cleans­ing and Stan­dard­iza­tion: Imple­ment auto­mat­ed tools to clean and stan­dard­ize data, ensur­ing it’s con­sis­tent and action­able.

Build a Bro­ker Lay­er: Cre­ate a bro­ker lay­er that stan­dard­izes data across sys­tems, enabling real-time data flow that can sup­port advanced pre­dic­tive mod­els.

Add a Pre­dic­tive Mod­el­ing Lay­er: Lever­age a ser­vice lay­er that taps into opti­mized and expert-dri­ven machine learn­ing (ML), neur­al net­works, and oth­er advanced pre­dic­tive AI mod­els. This lay­er can accu­rate­ly fore­cast demand and sup­ply by draw­ing from the cleaned, stan­dard­ized data along­side exter­nal ven­dor data, pro­duc­ing detailed pre­dic­tions.

The pre­dic­tive mod­el­ing lay­er out­puts data insights that can be fed direct­ly into an AI Agent, enabling it to act on real-time fore­casts and make deci­sions autonomous­ly. With this struc­tured foun­da­tion, star­tups can pro­gres­sive­ly scale an ecosys­tem for demand and sup­ply plan­ning, trans­form­ing com­plex and siloed data into a cohe­sive, pre­dic­tive real-time net­work.

The AI agent soft­ware ecosys­tem has devel­oped sig­nif­i­cant­ly in the past few months with progress in mem­o­ry, tool usage, secure exe­cu­tion, and deploy­ment. Below is an enter­prise archi­tec­ture dia­gram and a stan­dard enter­prise AI stack that can be used in SCI use cas­es.

SCI can, to a greater extent, use the stan­dard Agen­tic archi­tec­ture as a part of its AI Orches­tra­tion lay­er. The enter­prise appli­ca­tion stack dia­gram, drawn from a16z’s paper on LLM archi­tec­ture, is also adapt­able for oth­er trans­former-based use cas­es.

Com­pre­hen­sive AI Agents Stack and Mar­ket Map: A detailed overview of key lay­ers and capa­bil­i­ties for seam­less­ly adopt­ing and inte­grat­ing agen­tic solu­tions into SCI (source: a16z, Men­lo Ven­tures, Let­ta).

Why Now: Mov­ing from “No Phone” to a “Smart­phone”

Imag­ine try­ing to man­age your sup­ply chain with noth­ing but dis­parate spread­sheets and out­dat­ed reports. It is akin to try­ing to com­mu­ni­cate with­out a phone. The leap from today’s dis­joint­ed sys­tems to an AI-enabled, nat­ur­al lan­guage solu­tion is like going from no phone to an advanced smart­phone. Just as a smart­phone inte­grates voice, mes­sag­ing, inter­net, and apps into one seam­less expe­ri­ence, an AI-dri­ven plat­form would uni­fy inven­to­ry, pro­cure­ment, logis­tics, and real-time demand sens­ing into a sin­gle, acces­si­ble tool.

Such a sys­tem wouldn’t just offer insights—it would auto­mate key deci­sions and actions based on real-time data, empow­er­ing sup­ply chain man­agers with a lev­el of pre­ci­sion and agili­ty that feels as intu­itive as using a smart­phone. The shift is over­due, and the tools need­ed to make this trans­for­ma­tion are now with­in reach.

Ver­ti­cal AI: Indus­try-Spe­cif­ic Solu­tions as a Dif­fer­en­tia­tor

Ver­ti­cal AI will be instru­men­tal in tai­lor­ing solu­tions for spe­cif­ic indus­tries, where sup­ply chain dynam­ics and needs vary. For instance:

Food and Bev­er­age: The food indus­try has unique chal­lenges, includ­ing per­isha­bil­i­ty, sea­son­al­i­ty, and fluc­tu­at­ing con­sumer pref­er­ences. has demon­strat­ed its val­ue in address­ing indus­try-spe­cif­ic com­plex­i­ties by help­ing com­pa­nies to reduce food waste and opti­mize pro­duc­tion using real-time demand insights. This not only saves time and mon­ey, but also min­i­mizes food loss, cre­at­ing more effi­cient and sus­tain­able oper­a­tions.

Con­sumer Elec­tron­ics: The con­sumer elec­tron­ics sec­tor is char­ac­ter­ized by com­po­nent short­ages and rapid prod­uct cycles lead­ing to com­plex plan­ning needs. Ver­ti­cal AI mod­els designed for elec­tron­ics can fore­cast demand and sup­ply fluc­tu­a­tions, allow­ing man­u­fac­tur­ers to main­tain opti­mal inven­to­ry lev­els with­out over­stock­ing or risk­ing out-of-stock sit­u­a­tions.

By focus­ing on ver­ti­cal-spe­cif­ic AI, star­tups can pro­vide nuanced solu­tions that address the unique chal­lenges and pri­or­i­ties of each indus­try. This approach is essen­tial for devel­op­ing solu­tions that cre­ate real, sus­tain­able val­ue.

Rapid Growth and Mul­ti­ple Exit Oppor­tu­ni­ties

The SCI mar­ket is grow­ing at a record pace, dri­ven by its high frag­men­ta­tion and increased demand for real-time, inte­grat­ed solu­tions. Sup­ply chain soft­ware has trad­ed with low­er volatil­i­ty and at a high­er mul­ti­ple than SaaS com­pa­ra­bles across both pub­lic and pri­vate mar­kets. In 2023, pub­licly trad­ed sup­ply chain man­age­ment soft­ware saw a The M&A mar­ket has been active, with as com­pa­nies seek to build, acquire, or enhance their AI-dri­ven capa­bil­i­ties, accord­ing to data from Pitch­book.

Notable acqui­si­tions under­score this trend:

• project44 acquired Ocean Insight for $45 mil­lion, boost­ing its capa­bil­i­ties in ocean freight vis­i­bil­i­ty.
• Trim­ble Inc.’s $2.1 bil­lion acqui­si­tion of Trans­pore­on expand­ed its real-time logis­tics man­age­ment solu­tions.
• Coupa Soft­ware acquired LLa­ma­soft for $1.5 bil­lion, enhanc­ing its AI-dri­ven sup­ply chain plan­ning capa­bil­i­ties, which led to its sub­se­quent $8 bil­lion buy­out by Thoma Bra­vo in ear­ly 2023.

Investors see tremen­dous poten­tial in SCI,. Firms like Bain Cap­i­tal Ven­tures, which invest­ed $200 mil­lion in Ship­Bob in 2021, rec­og­nize the oppor­tu­ni­ty to cap­i­tal­ize on an indus­try under­go­ing rapid dig­i­tal trans­for­ma­tion. Giv­en that for crit­i­cal sup­ply chain process­es, the demand for auto­mat­ed, pro­duc­tized solu­tions is poised for sig­nif­i­cant growth. As Jaya Gup­ta and Ashu Garg from Foun­da­tion Cap­i­tal esti­mate, AI-enabled sup­ply chain solu­tions are tack­ling a $62 bil­lion mar­ket oppor­tu­ni­ty, with vast poten­tial across pro­cure­ment, demand plan­ning, and sup­pli­er intel­li­gence.

Recent M&A and spon­sor activ­i­ties by seg­ment.

For star­tups, this dynam­ic cre­ates mul­ti­ple path­ways to scale and exit. A robust SCI plat­form that can address inte­gra­tion and real-time plan­ning at scale has strong poten­tial for both acqui­si­tion by strate­gic play­ers and sub­stan­tial investor inter­est. The high demand for cohe­sive solu­tions that bridge dis­parate data sources, enhance sup­ply chain vis­i­bil­i­ty, and sup­port AI-dri­ven deci­sion-mak­ing, posi­tions a suc­cess­ful inte­gra­tion plat­form as a valu­able tar­get. Despite the chal­lenge, solu­tions that effec­tive­ly tack­le this prob­lem attract a broad array of buy­ers eager to deep­en their dig­i­tal capa­bil­i­ties and investors look­ing to cap­ture the mar­ket’s rapid growth.

How Star­tups Can Lead the SCI Trans­for­ma­tion

To rede­fine SCI, star­tups must focus on build­ing scal­able, pro­duc­tized AI solu­tions that uni­fy demand and sup­ply plan­ning in real time. Star­tups should pri­or­i­tize three crit­i­cal capa­bil­i­ties:

• End-to-End Vis­i­bil­i­ty: A uni­fied plat­form that con­sol­i­dates demand, inven­to­ry, pro­cure­ment, and logis­tics will deliv­er the holis­tic insights need­ed to opti­mize oper­a­tions.
• Real-Time Demand Sens­ing and Sup­ply Dis­rup­tion Detec­tion: AI-dri­ven demand sens­ing can empow­er com­pa­nies to adapt to changes as they occur, cre­at­ing a proac­tive approach to sup­ply chain man­age­ment.
• Auto­mat­ed Deci­sion-Mak­ing: Automat­ing crit­i­cal decisions—like adjust­ing pro­duc­tion sched­ules or rerout­ing ship­ments based on real-time data—will reduce man­u­al inter­ven­tion, stream­line process­es, and improve respon­sive­ness.

Long tail of hun­dreds of M&A and Agen­tic dis­rup­tion ripe com­pa­nies under $50M in invest­ed cap­i­tal.

The Time for AI-Dri­ven Sup­ply Chain Plan­ning Solu­tions is Now

The SCI mar­ket is ripe for trans­for­ma­tion. I believe con­sult­ing-led mod­els are los­ing rel­e­vance as com­pa­nies seek scal­able, pro­duc­tized solu­tions that enable real-time demand and sup­ply plan­ning. Star­tups have the chance to rede­fine the land­scape by attack­ing the core of the prob­lem, address­ing the “bel­ly of the beast” with AI-dri­ven solu­tions that inte­grate across sup­ply chain func­tions.

For those who act now, the oppor­tu­ni­ty is vast. The shift from frag­ment­ed, reac­tive plan­ning to a cohe­sive, AI-dri­ven, real-time ecosys­tem is akin to mov­ing from a world with no phones to one with smartphones—seamless, con­nect­ed, and always acces­si­ble. Now is the time for AI-first star­tups to lead this trans­for­ma­tion and ush­er in a new era of intel­li­gent, resilient sup­ply chains.

—

About the Author: Bivek Adhikari is an MBA Fel­low at Ground­Force Cap­i­tal. He is a tech investor with a strong data sci­ence back­ground, focus­ing on ear­ly-stage as well as Series B‑C com­pa­nies. He recent­ly ded­i­cat­ed over 4,000 hours to research­ing the state of sup­ply chain tech­nolo­gies. This involved engag­ing with start­up founders, incum­bent oper­a­tors, indus­try veterans/experts, emerg­ing growth stage com­pa­nies, invest­ment bankers exe­cut­ing M&A deals, con­sult­ing firms, prospec­tive ear­ly-stage, dis­tressed and defunct com­pa­nies as well as growth leads, fel­low investors track­ing the space and key play­ers, end users and cus­tomers from dif­fer­ent CPG brands, exec­u­tives from var­i­ous com­pa­nies, and inde­pen­dent con­sul­tants advis­ing For­tune 500 com­pa­nies, among oth­er stake­hold­ers.

If you are a builder who is cur­rent­ly work­ing on or inter­est­ed in build­ing in this space, he’d love to speak with you. He is also hap­py to chat if you are inter­est­ed in sup­ply chain tech­nolo­gies but unsure about white spaces. You can reach him at bivek@groundforcecapital.com.

Source:

⁽¹⁾Analy­sis based on a detailed mod­el built using data from over 30 sources, includ­ing Mar­ket­sand­Mar­kets, For­tune Busi­ness Insights, and Grand View Research.